ABSTRACT

   Although severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2)
   variants raise concerns about decreased vaccine efficacy, vaccines
   continue to confer robust protection in humans, implying that immunity
   beyond neutralization contributes to vaccine efficacy. In addition to
   neutralization, antibodies can mediate various Fc-dependent effector
   functions, including antibody-dependent cellular phagocytosis (ADCP),
   antibody-dependent neutrophil phagocytosis (ADNP) and
   antibody-dependent cellular cytotoxicity (ADCC). However, the specific
   role of each Fc-mediated effector function in contributing to COVID-19
   disease attenuation in human remains unclear. To fully define the
   potential immune correlates of Fc-mediated effector functions, we
   comprehensively analysed the above Fc-mediated effector functions in
   two study cohorts. In the CoronaVac vaccinee cohort, individuals
   without breakthrough infection exhibited higher levels of ADCP and ADNP
   activities with a greater degree of cross-reactivity compared to those
   who had breakthrough infection. A predictive model was established
   incorporating ADNP activity and IgG titre, achieving an area under the
   curve (AUC) of 0.837. In the COVID-19 patient cohort, BA.5-specific
   ADCP and ADNP responses were significantly reduced in COVID-19 patients
   with fatal outcomes compared to milder outcomes. The prognostic model
   incorporating WT, BA.5, and XBB.1.5 spike-specific ADNP demonstrated
   effective predictive ability, achieving an AUC of 0.890. Meanwhile,
   transcriptomic analysis of peripheral blood mononuclear cells (PBMCs)
   from COVID-19 patients in the acute phases of infection highlighted
   remarkably upregulation of neutrophil activity and phagocytic function,
   further reinforcing the essential role of ADNP. Collectively, our
   findings underscored Fc-mediated effector activities, especially
   neutrophil phagocytosis, as significant antibody biomarkers for the
   risk of SARS-CoV-2 breakthrough infection and COVID-19 prognosis.

   KEYWORDS: COVID-19, vaccine, SARS-CoV-2, antibody-dependent neutrophil
   phagocytosis, Fc effector function

Introduction

   The deployment of severe acute respiratory syndrome coronavirus 2
   (SARS-CoV-2) vaccines has significantly reduced morbidity and mortality
   worldwide [[41]1,[42]2]. However, due to waning immunity and the
   emergence of viral variants that evade neutralizing antibodies (NAbs)
   induced by vaccines targeting the wild-type (WT) SARS-CoV-2 strain
   [[43]3,[44]4], breakthrough infections caused by variants of concern
   (VOCs) have increased globally among fully vaccinated individuals
   [[45]5]. Nevertheless, in cases where antibodies remain detectable,
   vaccine-mediated protection against severe disease and death has
   remained robust [[46]6,[47]7]. While NAbs have been associated with
   protection against infection across various vaccine trials
   [[48]8,[49]9], overall antibody-binding titres also strongly correlate
   with protective immunity [[50]10–12]. Additionally, emerging evidence
   from animal models suggests that antibody Fc-effector functions,
   including opsonophagocytosis, are linked to immune protection following
   natural infection [[51]13,[52]14], convalescent plasma therapy [[53]15]
   and monoclonal therapy [[54]16]. Yet, whether Fc profiles are directly
   associated with vaccine-mediated protection in humans remains unclear.
   Understanding this relationship could provide critical insights for
   guiding booster strategies and the design of next-generation,
   variant-specific vaccines.

   Real-world studies have indicated that breakthrough infections caused
   by new variants are associated with substantially lower rates of severe
   disease and mortality [[55]17,[56]18] even when prior vaccinations or
   infections involved earlier SARS-CoV-2 strains. These observations
   indicated that other immune responses beyond neutralization play an
   important role in protection. Emerging evidence showed that antibody
   Fc-mediated effector functions induced by different vaccine platforms
   play an important role in immunoprotection [[57]19]. Indeed, immune
   complexes can be recognized by Fc receptors (FcRs) located on the
   surface of various immune cells, such as natural killer (NK) cells,
   macrophages, and neutrophils, facilitating antimicrobial activities
   that function beyond neutralization. These activities include processes
   such as antibody-dependent cellular phagocytosis (ADCP),
   antibody-dependent neutrophil phagocytosis (ADNP) and
   antibody-dependent cellular cytotoxicity (ADCC). Early enrichment of
   FcγR-binding antibodies targeting the S2 domain, rather than specific
   IgG antibody levels, was identified as a biomarker of survival in
   COVID-19 [[58]20,[59]21]. Additionally, vaccination across different
   platforms elicited significant levels of antibody Fc-mediated effector
   functions with substantial cross-reactivity [[60]19,[61]22,[62]23] and
   durability [[63]23,[64]24], which confer robust protection in
   situations where neutralization is compromised by viral escape
   mutations [[65]22,[66]25]. Fc-dependent effector functions have been
   shown to be crucial for the therapeutic effectiveness of monoclonal
   antibodies in both mice and hamsters [[67]23]. Similarly, phagocytic
   cells play a key role in antibody-driven viral clearance during
   SARS-CoV infection. The activation of alveolar macrophages and ADCP are
   critical for Fc-mediated antibody protection in a mouse model of
   influenza A virus (IAV) infection [[68]26,[69]27]. In addition, the
   alternative immune responses, such as T-cell-mediated responses, might
   also contribute to vaccine protection against SARS-CoV-2 by identifying
   and eliminating infected cells, thereby offering long-term immunity
   beyond neutralization [[70]28].

   Our previous study showed that the cross-reactive and durable Fc
   effector function of antibodies induced by vaccination with three-dose
   CoronaVac [[71]29]. Given the importance of Fc-effector functions, we
   aimed to comprehensively examine the potency of Fc-mediated effector
   functions, including ADCP, ADNP and ADCC, using cohorts of CoronaVac
   vaccine recipients and COVID-19 patients. In addition, we established
   two models using ADNP responses, which could effectively predict the
   risk of breakthrough infection and COVID-19 fatality. Our data
   highlighted that CoronaVac-induced antibody Fc-mediated effector
   functions, particularly ADNP activities, were strongly associated with
   the COVID-19 prevention and disease severity, which might serve as
   surrogate biomarkers for vaccine efficacy. Further studies will be
   performed to validate our prognosis models with ADNP activities,
   covering different vaccine platforms and emerging variants.

Materials and methods

Study cohort and sample collection

   We established two prospective study cohorts: the CoronaVac vaccinee
   cohort (cohort 1) and the COVID-19 patient cohort (cohort 2) in Nanjing
   Drum Tower Hospital, Nanjing, China ([72]Figure 1). Written informed
   consent was obtained from all participants before any study procedures
   were conducted. The study protocol was approved by the Nanjing Drum
   Tower Hospital Ethics Committee (2021-034-01).

Figure 1.

   [73]Figure 1.
   [74]Open in a new tab

   Study design. (a) In the CoronaVac vaccinee cohort, 149 participants
   who received three doses CoronaVac vaccine were enrolled, in which 101
   subjects experienced BA.5 breakthrough infection (BTI), while 48
   subjects were not breakthrough infected (non-BTI) within one month
   during the Omicron BA.5 wave. Blood samples were collected 12 months
   after the third CoronaVac dose and two weeks before exposure to the
   Omicron wave. (b) In the COVID-19 patient cohort, there were 30 mild
   non-hospitalized patients and 139 hospitalized patients, including 40
   patients with moderate disease, 54 patients with severe disease, and 45
   patients with fatal outcomes.

   In the CoronaVac vaccinee cohort, 149 healthcare workers who had
   received 3-dose CoronaVac vaccine (Sinovac Biotech, Beijing, China)
   were enrolled. Due to the lift of the COVID Zero policy, the outbreak
   of Omicron breakthrough infection offers us an opportunity to evaluate
   the role of antibody Fc-mediated effector functions in COVID-19 disease
   resolution. Serum samples were collected in early December 2022, 12
   months post last vaccination and prior to Omicron wave ([75]Figure
   1(a)). An online questionnaire was used to collect data on
   COVID-19-related symptoms and infection status during 16 December 2022
   and 16 January 2023. SARS-CoV-2 infection was confirmed by positive
   COVID-19 antigen or nucleic acid test in individuals who developed
   COVID-19-related symptoms. Based on breakthrough infection status, the
   149 vaccinees were divided into non-breakthrough infected (non-BTI)
   subjects (n = 48) and breakthrough infected (BTI) subjects (n = 101).

   In the COVID-19 patient cohort, a total of 169 patients were enrolled,
   and serum samples were collected 7–15 days after the onset of symptoms
   ([76]Figure 1(b)). These serum samples were retrieved from routine
   biochemical or immunological testing and stored at −80°C for future
   analysis. The COVID-19 patient cohort was further categorized into
   non-hospitalized patients (mild, n = 30) and hospitalized patients
   (n = 139). The hospitalized COVID-19 patients were further stratified
   into 3 subgroups based on their disease severity and clinical outcomes,
   including moderate cases (n = 40), severe cases (n = 54) and fatal
   cases (n = 45). The severity of COVID-19 was evaluated based on the
   ninth version of the Diagnosis and Treatment guidance of COVID-19
   [[77]22,[78]23]. Severe cases were defined as individuals meeting one
   of the following criteria: (1) Respiratory distress with respiratory
   rate over 30 per minute; (2) Hypoxia (oxygen saturation less than < 93%
   in the resting state); (3) Arterial blood oxygen partial pressure
   (PaO[2])/oxygen concentration (FiO[2]) less than 300 mm Hg; or (4)
   Severe disease complications including respiratory failure which
   requires mechanical ventilation, septic shock, or non-respiratory organ
   failure.

The measurement of SARS-CoV-2 surrogate neutralizing antibodies (NAbs)

   The measurement of surrogate NAbs specific to the WT strain was
   conducted using a one-step immune-competition assay, as previously
   described [[79]30] with the iFlash3000-C Chemiluminescence Immunoassay
   Analyzer (Shenzhen Yhlo Biotech Co., Ltd, China). This assay detects
   total antibody levels in serum that compete with the SARS-CoV-2 spike
   protein receptor binding domain (RBD) for binding to its receptor,
   angiotensin-converting enzyme II (ACE2). The negative reference range
   for SARS-CoV-2 surrogate NAbs was 0–10 IU/mL.

Serum SARS-CoV-2 spike-specific IgG antibody titre

   An in-house enzyme-linked immunosorbent test (ELISA) was used to
   evaluate antigen-specific serological antibodies against SARS-CoV-2
   [[80]29]. WT, BA.5 and XBB.1.5 Spike proteins (Vazyme, Cat# CG202,
   CG246, CG27, Nanjing, China) were coated onto high-binding 96-well
   plates at a concentration of 500 ng/mL and incubated overnight at 4°C.
   After incubation, the plates were washed five times with
   phosphate-buffered saline (PBS) containing 0.1% Tween-20. Then, the
   plates were blocked by adding PBS with 5% Bovine Serum Albumin (BSA)
   and 0.5% Tween-20 and incubated at 37°C for 1 h. After blocking, rewash
   the plates and add serum samples from vaccinated or infected
   individuals, diluted in PBS containing 1% BSA and 0.5% Tween-20. The
   initial dilution was 1:500, followed by a five-step serial dilution
   using a 3-fold gradient. Incubate the plates at room temperature for
   2 h. After incubation, the plates were washed five times, and
   horseradish peroxidase (HRP)-conjugated anti-IgG secondary antibody
   (Abcam, cat# ab6759, Cambridge, England) was added and incubated for
   1 h at room temperature. After the second incubation, the plates were
   rewashed, tapped dry, and developed with tetramethylbenzidine (TMB) for
   5 min. The reaction was stopped by adding 1M sulphuric acid. The
   optical density (OD) value was measured at 450 nm. The cut-off value
   was calculated by taking the average OD of 45 negative control healthy
   individuals. The greatest serum dilution that produced an OD value
   greater than the cut-off value of the healthy control group at the same
   dilution was used to determine the antibody endpoint titre. For the
   negative controls, the samples were collected from healthy individuals
   aged 18 to 60 years, with a similar gender ratio, prior to 2019. These
   individuals had neither been infected with SARS-CoV-2 nor received a
   COVID-19 vaccine. The demographic information of the healthy control
   group was summarized in Supplemental Table 4.

Fc-mediated effector functional assays

   ADNP and ADCP assays were carried out as previously described
   [[81]29,[82]31]. All experiments were performed in triplicates to
   ensure the consistency of our results. To prepare the biotinylated
   SARS-CoV-2 spike protein, antigens were biotinylated using Sulfo-NHS-LC
   biotin (Thermo Fisher Scientific, cat# A39257, MA, USA) following the
   manufacturer’s protocol. Excess biotin was removed with a Zeba Spin
   desalting column (Thermo Fisher Scientific, cat# A44300, MA, USA). The
   biotinylated antigens were then mixed with 1 μm yellow-green
   fluorescent NeutrAvidin beads (Invitrogen, cat# F8776, MA, USA) in a
   1:1 ratio and incubated overnight at 4°C in the dark. After incubation,
   the beads were washed with PBS at 16,000 g for 15 min and resuspended
   in PBS containing 0.5% BSA.

   For the ADNP assay, neutrophils were isolated from the whole blood of
   three healthy adults using a lysis method. In details, the serum was
   diluted 25-fold in sterile PBS containing 0.5% Tween-20 and 1% BSA, and
   incubated with antigen-coupled beads at 37°C for 2 h to form immune
   complexes. After incubation, sterile PBS was used to wash away unbound
   immunoglobulins. PBMCs were resuspended at a concentration of 2.5 × 10⁵
   cells/ml in R10 media. A total of 25,000 cells per well were added to
   each well containing immune complexes, mixed by shaking and incubated
   for 1 h at 37°C, 5% CO[2]. After incubation, the PBMCs were washed with
   PBS and stained for CD66b + cells using APC-conjugated anti-CD66b
   antibody (BioLegend, cat# 17-0666-42, CA, USA) to identify neutrophils.
   The cells were then fixed with 4% paraformaldehyde (PFA) (Leagene, cat#
   DF0135, Beijing, China) for 30 min, followed by flow cytometry
   analysis. A blank control containing only beads and neutrophils was
   included as a negative control.

   The ADCP assay was performed as previously described [[83]29]. The
   diluted serum was used to prepare immune complexes following the same
   method described above. After washing to remove unbound immunoglobulins
   with sterile PBS, 25,000 THP-1 cells per well were added in R10 medium,
   mixed by shaking to ensure even distribution and incubated at 37°C with
   5% CO[2] for 1 h. Following incubation, THP-1 cells were washed with
   PBS and then fixed with 4% PFA (Leagene, cat# DF0135, Beijing, China)
   for 30 min, followed by flow cytometry analysis. A blank control
   containing only beads and THP-1 cells was included as a negative
   control. The geometric mean fluorescent intensity (gMFI) of the
   phagocytosed beads (bead + THP-1 or neutrophils, classified as positive
   cells) (Figure S1) and the percentage of cells that had phagocytosed
   beads were determined using flow cytometry. Phagoscore = % positive
   cells × geometric mean fluorescent intensity of positive
   cells/1000,000. The corrected phagoscore was defined as the actual
   phagoscore derived from the sample minus phagoscore score derived from
   the blank control. If the phagoscore calculated was below zero, it
   would be set as zero. The BD Accuri^TM C6 Plus Flow Cytometer was used
   for flow cytometry, and FlowJo V10.8.1 was used for analysis.

   To assess the ADCC activity of serum antibodies, shake-cultured 293F
   cells stably expressing the spike protein of the original strain served
   as target cells. While Jurkat-FcγRIII-NFAT-Luc reporter cells (Vazyme,
   cat# DD1301-1, Nanjing, China) were used as effector cells. These cells
   contain a functional NFAT transcription factor, which regulates the
   expression of the Lucia luciferase gene. The Jurkat-FcγRIII-NFAT-Luc
   cells stably express the FcγRIIIa (CD16) receptor. When antibodies bind
   to the CD16 receptor, the NFAT pathway is activated, causing NFAT to
   translocate to the nucleus and bind to the promoter of the Lucia
   luciferase gene, thereby driving its expression.

   The ADCC levels were detected using luciferase activity assays as
   previously described [[84]32]. In details, serum samples diluted 1:60
   in R10 medium were incubated with 25, 000 target 293F cells per well at
   37°C and 5% CO₂ for 1 h. After incubation, 75, 000 Jurkat cells were
   added to each well, gently mixed, and incubated for 12 h under the same
   conditions. Following the incubation, luciferase substrate (Vazyme,
   cat# DD1204-03, Nanjing, China) was added, and the wells were shaken to
   ensure full reaction. Relative light units (RLU) were measured
   according to the PerkinElmer (Waltham, MA) instruction manual. Sera
   samples from healthy archived persons from 2019 were used as the
   unexposed and unvaccinated donors for the negative control experiment.
   ADCC was measured as the fold induction of Luciferase activity compared
   to the negative control sera.

PBMC transcriptomic analysis

   PBMCs from forty volunteers were included in the transcriptomic
   analysis, collected 12 months after receiving three doses of the
   CoronaVac vaccine. Among them, twenty developed COVID-19-related
   symptoms and tested positive for SARS-CoV-2 antigen or nucleic acid
   within five days after sample collection, referring vaccinees during
   the acute phase of infection. As a negative control, twenty vaccinees
   who neither developed COVID-19 symptoms nor tested positive for
   SARS-CoV-2 antigen or nucleic acid were included as vaccinees without
   infection group.

   Total RNA was prepared and used for the RNA-seq library construction,
   and subsequently sequenced on an Illumina Novaseq platform and 150 bp
   paired-end reads were generated [[85]33]. The human genome version of
   hg38 was used as reference. FeatureCounts vl.5.0-p3 was used to count
   the reads numbers mapped to each gene. Differential expression analysis
   of two groups was performed using the DESeq2R package (v 1.20.0). The
   resulting p values were adjusted using Benjamini and Hochberg's
   approach to control the false discovery rate (FDR). Genes with an
   adjusted p values < 0.05 found by DESeq2 were assigned as
   differentially expressed. Gene Ontology (GO) and Kyoto Encyclopedia of
   Genes and Genomes (KEGG) pathway enrichment analysis were performed
   using clusterProfiler package in R software and corrected p values less
   than 0.05 were considered significantly enriched by differentially
   expressed genes.

Statistical analysis

   Single-variable comparisons between groups were performed using the
   Mann–Whitney U test. Multiple comparisons of ADCP, ADNP and ADCC
   responses were conducted using the Kruskal–Wallis test with FDR
   correction. Data were analysed using GraphPad Prism (version 9.0.1, La
   Jolla, California, USA). All statistical analyses were performed in a
   two-sided manner, and p values < 0.05 was considered statistically
   significant. * indicates p < 0.05, ** indicates p < 0.01, *** indicates
   p < 0.001, **** indicates p < 0.0001, and ns indicates no significant
   difference.

   A partial least square discriminant analysis (PLS-DA) was used for the
   classification and visualization of immune parameters between the two
   groups. Selected immune features were ranked based on the Variable
   Importance in Projection (VIP) score and the first two latent variables
   (LVs) of the PLS-DA model were used for visualization. R package
   “ropls” (version 1.20.043) and “glmnet” (version 4.0.244) and the 27
   systemseRology R package (version 1.1)
   ([86]https://github.com/LoosC/systemsseRology) were used for analysis.

   The predictive ability of the immunological index was evaluated using a
   receiver operating characteristic (ROC) curve. Variables based on the
   VIP score were selected for the multivariate models. A binary logistic
   regression analysis was used to combine multiple predictors. The area
   under the ROC curve (AUROC) was calculated. Data analysis was performed
   using Medcalc version 22.009 software (MedCalc Software Ltd, Belgium).

Results

Robust Fc-mediated phagocytosis activities induced by CoronaVac in
non-breakthrough infected groups

   To evaluate humoral correlates of immunity against COVID-19, we
   performed a detailed profiling of the humoral immune response to WT,
   BA.5 and XBB.1.5 of SARS-CoV-2, including NAbs, spike-specific IgG
   titres, ADCP, ADNP and ADCC. Here, we characterized the antiviral
   effector function of antibodies in serum by assessing antibody-mediated
   phagocytosis in a prospective cohort of SARS-CoV-2 naïve healthcare
   workers ([87]Figure 1(a) and Supplemental Table 1).

   Although previous studies have confirmed the association between
   antibody-binding titres, neutralization and protection in phase 3
   COVID-19 vaccine trials [[88]10,[89]34], comparable levels of IgG
   titres specific to WT and BA.5 spike proteins were observed between BTI
   patients and non-BTI patients. In contrast, serum samples from the
   non-BTI group exhibited significantly higher level of XBB.1.5
   spike-specific IgG titres compared to the BTI group (p = 0.0002)
   ([90]Figure 2(a)). Interestingly, there was a trend indicating that the
   non-BTI group displayed a higher degree of cross-reactivity with the
   XBB.1.5 spike, but not with the BA.5 spike, compared to the BTI group
   ([91]Figure 2(b)). Additionally, compared to BTI individuals, non-BTI
   individuals had higher levels of ADCP activities specific to WT, BA.5
   and XBB.1.5 spike proteins (p = 0.0001, 0.0013 and 0.0069 respectively)
   ([92]Figure 2(c)). However, sera from the BTI group exhibited slightly
   better cross-reactive ADCP activity against the XBB.1.5 spike protein
   compared to the non-BTI group ([93]Figure 2(d)). Moreover, the non-BTI
   group demonstrated a greater magnitude and higher degree of
   cross-reactive XBB.1.5 spike-specific ADNP activity compared to the BTI
   group (p < 0.0001) ([94]Figure 2(e,f)). Although minimal levels of ADCC
   activity were detected across the cohorts, sera from the non-BTI group
   exhibited higher ADCC activities compared to the BTI group (p = 0.0361)
   ([95]Figure 2(g)). In contrast, comparable levels of NAbs were observed
   between BTI and non-BTI group, possibly due to the relative low level
   of serum neutralization activities which could not confer disease
   prevention [[96]11] ([97]Figure 2(h)). Our data suggested that
   Fc-mediated effector functions were remarkably higher in non-BTI group
   than BTI group.

Figure 2.

   [98]Figure 2.
   [99]Open in a new tab

   Baseline humoral immune responses between the vaccinees with
   breakthrough infection (BTI) and vaccinees without breakthrough
   infection (non-BTI). (a) The serum IgG, (c) antibody-dependent cellular
   phagocytosis (ADCP) and (e) antibody-dependent neutrophil phagocytosis
   (ADNP) specific to spike protein of WT, BA.5, and XBB.1.5 between the
   BTI and the non-BTI. Fold changes in (b) IgG titre, (d) ADCP and (f)
   ADNP responses specific to the Omicron BA.5 and XBB.1.5 subvariants
   spike proteins relative to those specific to the WT strain. The dotted
   line represents no fold change, and the solid lines indicate the median
   value. (g) WT spike-specific antibody-dependent cellular cytotoxicity
   (ADCC) response between the BTI and non-BTI groups. The Mann–Whitney U
   test was used for comparison between the two groups. * indicates
   p < 0.05, ** indicates p < 0.01, *** indicates p < 0.001, ****
   indicates p < 0.0001. ns indicates no significant difference.

Vaccine-elicited Fc effector biomarkers that differentiate BTI and non-BTI
individuals

   Given the notable differences in binding antibody titres and Fc
   effector functions between the BTI and non-BTI groups, we aimed to
   determine whether Fc effector parameters might serve as potential
   biomarkers to differentiate BTI. To this end, a PLS-DA was performed,
   incorporating the tested humoral responses as well as sex and age
   across the two groups ([100]Figure 3(a)). This analysis distinguished
   BTI from non-BTI samples, indicating significant differences in humoral
   responses between the two groups. The top three biomarkers identified
   from VIP scores between the BTI and non-BTI groups were BA.5
   spike-specific IgG, XBB.1.5 spike-specific ADNP and WT spike-specific
   ADNP ([101]Figure 3(b)). Given that these three biomarkers contribute
   most significantly to differentiating between BTI and non-BTI groups,
   the top three markers were adopted to establish a predictive model
   using logistic regression to assess the potential breakthrough
   infection risk. The predictive model equation is as follows:
   P[BTI] = 2.88776–0.085186*XBB.1.5 spike ADNP + 0.000068797*BA.5 spike
   IgG + 0.022264*WT spike ADNP.

Figure 3.

   [102]Figure 3.
   [103]Open in a new tab

   Selected Fc-mediated function biomarkers and development of the risk
   prediction model for breakthrough infection. (a) A partial least
   squares discriminant analysis (PLS-DA) comparing BTI and non-BTI
   groups. (b) The top three indicators with the highest Variable
   Importance in Projection (VIP) scores, including BA.5 spike-specific
   IgG, WT spike-specific ADNP and XBB.1.5 spike-specific ADNP, were
   selected based on their VIP scores, which measure the contribution of
   each variable to the classification model in PLS-DA distinguishing
   between BTI and non-BTI groups. (c) Receiver operating characteristic
   (ROC) curves of the combined predictive model, using three biomarkers
   with the highest VIP scores, alongside the 3 individual top-performing
   indicators. (d) Area under the receiver operating characteristic curves
   (AUROCs) for eleven individual tested immune biomarker, as well as for
   the predictive model. The AUROC values provide a quantitative
   assessment of the model’s performance, highlighting its capacity to
   accurately classify individuals with breakthrough infection status. The
   vertical dotted line represents AUROC = 0.6. Curves with p values less
   than 0.05 are shown in red, while those with p values greater than 0.05
   are shown in grey.

   To validate the performance of this predictive model, we generated
   receiver operating characteristic (ROC) plots to compare the predictive
   ability of each humoral antibody parameter. BA.5 spike-specific IgG, WT
   spike-specific ADNP, and WT spike-specific IgG yielded the modest
   predictive ability with AUC of 0.548, 0.552 and 0.552, respectively
   ([104]Figure 3(c,d)). To further assess the effectiveness of the
   combined parameters in predicting the risk of breakthrough infection,
   we applied the 3-indicator model to distinguish between BTI and non-BTI
   group. The AUROC of the model was 0.837 (95% confidence interval (CI):
   0.767–0.892, sensitivity: 77.00%, specificity: 77.08%), demonstrating
   superior discriminatory ability in identifying COVID-19 patients with
   breakthrough infection ([105]Figure 3(c,d)).

   To further validate our findings, quartile ranges were set to establish
   the risk of breakthrough infection based on the magnitude of ADCP and
   ADNP activities specific to WT, BA.5 and XBB.1.5 spike protein
   ([106]Table 1). Specifically, participants with WT spike-specific ADCP
   responses in the third (Phagoscore: 145.23–225.37) and fourth
   (226.28–414.12) quartiles had 78.8% (p = 0.008, OR = 0.212,
   95%CI:0.067–0.670) and 83.2% (p = 0.002, OR = 0.168, 95%CI:0.054–0.524)
   smaller odds respectively than those in the first quartile.
   Participants with XBB.1.5 spike-specific ADNP responses in the third
   (48.74–69.25) and fourth (71.21–168.50) quartiles had 82.4% (p = 0.013,
   OR = 0.176, 95%CI:0.045–0.694) and 94.9% (p < 0.0001, OR = 0.051,
   95%CI:0.013–0.199) smaller odds respectively than those in the first
   quartile. Collectively, our data suggested that the high magnitude of
   Fc activities was associated with a relatively lower risk of
   breakthrough infection.

Table 1.

   Risk of Omicron BA.5 variant infection within 1 month of blood
   sampling.
   Fc effector functions Quartile Phagocytosis score Infected Infection
   rate (%)
   WT spike ADCP Q1 (0–25%) <77.25 32/37 86.49
   Q2 (25–50%) 79.25–144.51 27/37 72.97
   Q3 (50–75%) 145.23–225.37 22/37 59.46
   Q4 (75–100%) 226.28–414.12 19/37 51.35
   BA.5 spike ADCP Q1 (0–25%) <16.39 28/37 75.68
   Q2 (25–50%) 16.64–32.56 32/37 86.49
   Q3 (50–75%) 32.79–67.36 21/37 56.76
   Q4 (75–100%) 67.87–200.45 19/37 51.35
   XBB.1.5 spike ADCP Q1 (0–25%) <24.85 31/37 83.78
   Q2 (25–50%) 24.96–41.96 26/37 70.27
   Q3 (50–75%) 42.15–65.79 22/37 59.46
   Q4 (75–100%) 65.90–252.01 21/37 56.76
   WT spike ADNP Q1 (0–25%) <59.74 25/37 67.57
   Q2 (25–50%) 60.33–89.81 28/37 75.68
   Q3 (50–75%) 90.13–134.27 27/37 72.97
   Q4 (75–100%) 138.74–359.46 20/37 54.05
   BA.5 spike ADNP Q1 (0–25%) <55.76 26/37 70.27
   Q2 (25–50%) 56.67–79.43 29/37 78.38
   Q3 (50–75%) 80.58–106.32 25/37 67.57
   Q4 (75–100%) 107.44–232.67 20/37 54.05
   XBB.1.5 spike ADNP Q1 (0–25%) <30.12 33/37 89.19
   Q2 (25–50%) 31.07–48.69 29/37 78.38
   Q3 (50–75%) 48.74–69.25 24/37 64.86
   Q4 (75–100%) 71.21–168.50 14/37 37.84
   [107]Open in a new tab

Diminished Fc-mediated phagocytosis responses observed in COVID-19 patients
with fatal outcomes compared to those with non-fatal outcomes

   To define the differences in Fc-mediated phagocytosis among COVID-19
   patients and assess correlations with disease severity, we further
   analysed the Fc-mediated effector activities across patients with
   different clinical outcomes, including mild, moderate, severe and fatal
   cases ([108]Figure 1(b) and Supplemental Table 2). We found that
   non-hospitalized patients exhibited a higher magnitude of ADCP
   responses against the BA.5 subvariant compared to hospitalized patients
   (p < 0.0001), whereas no such effect was observed for the WT spike or
   XBB.1.5 spike ([109]Figure 4(a)). Specifically, COVID-19 patients with
   mild disease exhibited consistently higher levels of ADCP responses
   specific to WT, BA.5 and XBB.1.5 spike than patients with fatal
   outcomes (p = 0.018, p < 0.0001, and p = 0.0021 respectively).
   Furthermore, patients with severe COVID-19 showed elevated BA.5
   spike-specific ADCP responses compared to those with moderate disease
   and fatal outcomes (p = 0.0361 and p = 0.002, respectively). Patients
   with fatal outcomes demonstrated significantly diminished levels of
   XBB.1.5 spike-specific ADCP responses compared to both with severe and
   moderate disease (p = 0.0021 and p = 0.0020, respectively). Different
   from ADCP, ADNP responses specific to WT spike, BA.5 spike and XBB.1.5
   spike proteins were remarkably elevated in patients with mild disease
   compared to hospitalized patients (p = 0.0115, p < 0.0001, and
   p = 0.0006 respectively) ([110]Figure 4(b)). Specifically, BA.5
   spike-specific ADNP responses were significantly higher in patients
   with mild disease than in those with severe and fatal disease
   (p < 0.0001, p < 0.0001, respectively), but not significantly different
   from patients with moderate disease (p = 0.3764). A similar trend was
   also observed for XBB.1.5 spike-specific ADNP responses. In addition,
   ADCC activities were significantly elevated in COVID-19 patients with
   mild disease, compared to hospitalized patients (p < 0.0001) and across
   all subgroups (p < 0.0001, p = 0.0002, and p < 0.0001, respectively)
   ([111]Figure 4(c)). A similar pattern was observed for patients with
   mild disease who had significantly higher NAbs compared to hospitalized
   patients (p < 0.0001) ([112]Figure 4(d)). However, there were no
   notable differences in either ADCC activities or NAb titres were
   observed among patients with moderate, severe, and fatal disease
   ([113]Figure 4(c,d)).

Figure 4.

   [114]Figure 4.
   [115]Open in a new tab

   ADCP and ADNP responses among COVID-19 patients with different clinical
   outcomes. (a) ADCP and (b) ADNP specific to the WT, BA.5 and XBB.1.5
   spike proteins among COVID-19 patients with varying clinical outcomes.
   (c) ADCC and (d) neutralizing antibody titres specific to the WT spike
   protein in COVID-19 patients with different clinical outcomes. (e-f)
   Cross-reactivity of ADCP and ADNP to WT, BA.5 and XBB.1.5 spike
   proteins in patients with different clinical outcomes, categorized as
   mild, moderate, severe, and fatal groups. The Mann–Whitney U test was
   used for two-group comparisons. While the Kruskal–Wallis test with FDR
   correction was applied for comparisons across four groups. * indicates
   p < 0.05, ** indicates p < 0.01, *** indicates p < 0.001, ****
   indicates p < 0.0001. ns indicates no significant difference.

   The cross-reactivity of serum ADNP and ADCP activities was also
   assessed. Serum samples from patients with mild disease exhibited
   higher cross-reactivity in BA.5 spike-specific ADCP responses compared
   to those from patients with moderate disease (p = 0.0059). Meanwhile,
   serum from patients with fatal outcomes demonstrated greater
   cross-reactivity in BA.5 spike (p = 0.045 and 0.0002) and XBB.1.5
   spike-specific ADCP (p = 0.0065 and 0.0065) responses compared to
   patients with severe and moderate disease ([116]Figure 4(e)).
   Interestingly, patients with fatal outcomes exhibited significantly
   reduced cross-reactivity in BA.5 spike-specific ADNP responses compared
   to those with severe and moderate disease (p = 0.0296 and 0.0191).
   Additionally, fatal COVID-19 patients showed markedly lower
   cross-reactivity in XBB.1.5 spike-specific ADNP responses compared to
   patients with severe, moderate and mild disease (p = 0.0293, 0.0001 and
   0.0017) ([117]Figure 4(f)). Overall, serum from patients with mild
   disease displayed robust Fc-mediated effector functions, whereas
   patients with fatal outcomes exhibited only modest Fc-mediated effector
   functions, with a skew towards BA.5 spike-specific responses.

ADNP accurately predicted the clinical outcomes in COVID-19 patients

   To determine whether Fc-mediated phagocytosis could serve as a
   biomarker to predict the disease progression of COVID-19, we further
   divided the hospitalized COVID-19 patients into two groups, including
   patients with fatal and non-fatal (including moderate and severe cases)
   outcomes. PLS-DA analysis was conducted to evaluate whether overall
   Fc-mediated effector functions differ between the two groups. As shown
   in [118]Figure 5(a), a partial separation between non-fatal and fatal
   COVID-19 patients was observed. Notably, the top 3 VIP scores from the
   PLS-DA analysis indicated that patients with non-fatal outcomes
   exhibited selective enrichment in XBB.1.5 spike-specific ADNP, WT
   spike-specific ADNP, and BA.5 spike-specific ADNP responses in their
   immune profiles ([119]Figure 5(b)). Similarly, we combined the three
   biomarkers into a predictive model to assess fatal outcomes. The
   prognosis model equation is as follows:
   P[fatality] = 0.95858–0.096526*XBB.1.5 spike ADNP +0.024837*BA.5 spike
   ADNP +0.0060758* WT spike ADNP. The AUROCs for predicting COVID-19
   fatality were 0.851 (95% CI: 0.778–0.908), 0.807 (95% CI: 0.731–0.870),
   0.708 (95% CI: 0.624–0.782) using XBB.1.5 spike-specific ADNP, WT
   spike-specific ADCP and XBB.1.5 spike-specific ADCP, respectively
   ([120]Figure 5(c)). Using the combined parameters with the top 3 VIP
   scores, the model achieved an AUC of 0.890 (95% CI:0.821–0.939), with
   80.65% sensitivity and 82.80% specificity ([121]Figure 6(d)). Overall,
   the model incorporating the specific ADNP responses demonstrated strong
   performance in identifying the fatal outcomes of COVID-19 patients.

Figure 5.

   [122]Figure 5.
   [123]Open in a new tab

   Selected Fc-mediated function biomarkers and the development of
   clinical prognosis model for COVID-19 fatality. (a) A PLS-DA was
   conducted to compare non-fatal versus fatal COVID-19 patients. (b) The
   top three indicators with the highest VIP scores, including WT, BA.5
   and XBB.1.5 spike-specific ADNP, were selected based on their VIP
   scores, which measure the contribution of each variable to the
   classification model in PLS-DA distinguishing between fatal and
   non-fatal groups. (c) ROC curves of the prognosis model, using three
   biomarkers with the highest VIP scores, alongside the 3 individual top
   indicators. (d) AUROCs for 9 individual tested immune biomarker, as
   well as for the prognosis model. The AUROC values provide a
   quantitative assessment of the model’s performance, highlighting its
   capacity to accurately classify individuals with fatal COVID-19
   outcomes. The vertical dotted line represents AUC = 0.6. Curves with p
   values less than 0.05 are shown in red, while those with p values
   greater than 0.05 are shown in grey.

Figure 6.

   [124]Figure 6.
   [125]Open in a new tab

   Transcriptomic analysis of PBMCs from CoronaVac vaccinees during the
   acute phase of SARS-CoV-2 breakthrough infection and those without
   infection. (a) t-Distributed stochastic neighbour embedding (t-SNE)
   plot of PBMC transcriptomic data from twenty vaccinees in the acute
   phase of SARS-CoV-2 infection and twenty uninfected vaccinees. (b)
   Volcano plot showing differentially expressed genes in PBMCs of
   vaccinees at the acute phase of infection compared to uninfected
   vaccinees. The selection criteria are padj < 0.05 and an absolute
   log[2]fold change > 1 (c) Gene ontology (GO) and (d) Kyoto Encyclopedia
   of Genes and Genomes (KEGG) functional enrichment analysis of
   upregulated genes in vaccinees during the acute phase of infection
   compared to uninfected vaccinees. (e) Heatmap of significant
   differentially expressed genes comparing PBMC from vaccinees during the
   acute phase of SARS-CoV-2 infection versus vaccinees without
   breakthrough infection.

Transcriptomic analysis revealed neutrophil activation in vaccinees during
the acute phase of infection

   To further validate the protective role of Fc-mediated effector
   function in disease prevention and progression, we performed a
   transcriptomic analysis comparing blood samples from twenty vaccinees
   during the acute phase of breakthrough infection with twenty vaccinees
   without breakthrough infection ([126]Figure 6). The clinical and
   demographic information of the enrolled subjects was summarized in
   Supplemental Table 3. The principal component analysis showed that
   targeted transcriptomic profiles could effectively distinguish patients
   undergoing acute SARS-CoV-2 infection from uninfected vaccinees
   ([127]Figure 6(a)). 2209 genes were significantly upregulated whereas
   4436 genes were downregulated in vaccinees in the acute phase of
   SARS-CoV-2 infection ([128]Figure 6(b)). Gene ontology (GO) analysis
   showed that the most strongly regulated biological process categories
   were associated with neutrophil activation, neutrophil activation
   involved in immune response, neutrophil degranulation and phagocytosis
   ([129]Figure 6(c)). Kyoto Encyclopedia of Genes and Genomes (KEGG)
   pathway analysis of gene expression profiles indicated a notable
   upregulation in Fcγ receptor-mediated phagocytosis, chemokine
   signalling pathway, neutrophil extracellular trap (NET) formation, and
   B cell receptor (BCR) signalling pathway ([130]Figure 6(d)). As shown
   in [131]Figure 6(e), compared to PBMCs from healthy control.
   Transcriptomic analysis of PBMCs from individuals in the acute phase of
   infection revealed upregulated gene expression related to neutrophil
   activation and regulation of neutrophil immune responses (TCN2, LILRA2,
   CD14, ANXA5, LILRB4, CCR1, CFD, BST2, FPR2, MNDA, MS4A3, SERPINB10,
   LRG1, C3AR1, PRTN3, LTF, GPR84, CLEC5A), Fcγ receptor-mediated
   phagocytosis (SYK, HCK, FCGR1A, LYN, FCGR2A, FCGR1CP, MARCKS, MYO10,
   DOCK1, FCGR3B) and B cell responses (FOS, LILRA1, LILRA6, BTK, IFITM2,
   LILRA5). Collectively, our data suggest that Fc-mediated neutrophil
   phagocytosis might play an essential role in the acute phase of
   infection and pathogen clearance, which may be associated with
   different disease outcomes.

Discussion

   Neutralizing antibody responses are often employed to estimate
   protective efficacy in the context of COVID-19 vaccination [[132]35].
   However, it is worth mentioning that even when neutralization is
   compromised in the setting of the SARS-CoV-2 variants and breakthrough
   infection, vaccine-induced protection against hospitalization remains
   relatively high [[133]36–38]. Antibodies that activate innate immune
   responses can both directly facilitate the rapid clearance of viral
   particles and eliminate infected cells by engaging the immune system
   [[134]39], suggesting that immune mechanisms beyond antibody-mediated
   virus neutralization may play an essential role in protecting against
   severe disease.

   There is increasing evidence suggesting that beyond neutralization,
   vaccine-induced antibodies are also capable of mediating an array of
   antibody effector functions that have been linked to protection against
   severe disease and death [[135]23] and monoclonal therapeutic activity
   [[136]24,[137]40]. Nevertheless, the protective role of Fc effector
   functions in preventing SARS-CoV-2 infection and limiting COVID-19
   severity has not been well established. Previously, we profiled the Fc
   effector functions including ADCP and ADNP responses induced by 3-dose
   CoronaVac, in which Fc-mediated effector phagocytosis was
   cross-reactive with Omicron variants and long-lasting [[138]29], which
   provide a timely and comprehensive insight of the mechanism of vaccine
   efficacy. In contrast to NAbs, which are constrained to target a finite
   number of regions on the spike protein pertinent to attachment or
   fusion, antibodies mediating Fc-activity possess the capability to bind
   extensively across the entire antigenic surface. Fc-activity
   fundamentally necessitates the formation of immune complexes and the
   strategic arrangement of antibodies featuring Fc-domains that are
   readily accessible to local immune cells. Despite a diminished yet
   persistent binding affinity towards the Omicron spike antigen, IgG
   antibodies may persist in opsonizing both the virus and virally
   infected cells. Consequently, non-neutralizing antibodies that harness
   the potential of Fc-biology could play a pivotal role in expediting the
   elimination of infected cells immediately following the onset of
   infection.

   Here we further expanded our understanding of Fc-mediated effector
   functions elicited by CoronaVac, which might contribute to disease
   protection and clinical outcomes of SARS-CoV-2 infection. In our first
   cohort, since most participants received the last vaccination 12 months
   ago, the minimum level of serum neutralization activities was detected
   due to a pronounced loss of antibodies and the immune escape of Omicron
   variants. This also offered us a great opportunity to determine the
   preventive role of Fc-mediated activities to prevent the infection and
   disease control. We developed a model for predicting risk of BA.5
   breakthrough infection which incorporated three parameters including
   two ADNP parameters (ADNP specific to XBB.1.5 and WT spike) and BA.5
   spike IgG, achieved an AUC of 0.837 for predicting the risk of BA.5
   breakthrough infection, with 77.08% specificity and 77.00% sensitivity.
   Meanwhile, the three-ADNP-indicator clinical prognosis model was
   established to distinguish COVID-19 patients with fatal and non-fatal
   outcomes achieved an AUC of 0.890, with 82.80% specificity and 80.65%
   sensitivity. For the first time, our models highlight the importance of
   Fc-mediated effector functions especially neutrophil-mediated
   phagocytosis that may contribute to the prevention of SARS-CoV-2
   infection as well as disease control. Therefore, those at high risk of
   infection will be encouraged to boost with the updated COVID-19
   vaccine. For individuals at high risk of COVID-19-related fatality,
   proactive clinical interventions such as the prompt administration of
   antiviral medications are strongly recommended.

   Neutrophils are the most abundant circulating leukocyte in the blood
   and can rapidly migrate to sites of infection and can mediate a range
   of effector responses [[139]41,[140]42]. FcγRs allow neutrophils to
   interact with virus particles and infected cells opsonized by specific
   IgG antibodies, leading to neutrophil activation, degranulation,
   phagocytosis, and the induction of neutrophil extracellular traps
   (NETs). Neutrophils express antibody receptors for IgG, including the
   activating high-affinity FcγRI and the low-affinity FcγRIIa and
   FcγRIIIb, as well as IgA, which binds to the low-affinity FcαRI.
   FcγRIIIb is the most abundantly expressed receptor on neutrophils and
   is exclusively expressed on these cells. which specifically binds to
   IgG1 and IgG3. In comparison, FcγRIIa is expressed at lower levels than
   FcγRIIIb on neutrophils, and similarly binds to IgG1 and IgG3. Although
   FcγRI is expressed at lower levels, it can bind IgG1, IgG3, and IgG4,
   making it a high-affinity receptor. These receptors distinguish
   neutrophils from their innate immune functions by endowing them with
   the ability to mount antigen-specific immune responses.

   In the context of COVID-19, the presence of FcγRIIIb-binding antibodies
   was found to be enriched in individuals who were protected against
   COVID-19 breakthrough, linked to neutrophil-mediated immune complex
   clearance [[141]43]. The protection against SARS-CoV-2 BA.5 infection
   in the upper respiratory airways correlated with binding, neutralizing,
   and ADNP activities of the serum antibody elicited by COVID-19 vaccine
   [[142]44]. Meanwhile, impairment of neutrophil functions, homeostasis,
   and inflammation was associated with COVID-19 disease severity
   [[143]45] and post-COVID-19 pulmonary sequelae [[144]39]. Similar
   mechanisms were also observed in alternative infectious diseases.
   Additionally, antibody specific for soluble glycoprotein (sGP) drives
   neutrophil-mediated phagocytosis and could predict vaccine-mediated
   protection. In Ebola vaccine, protective sGP-specific monoclonal
   antibodies have elevated neutrophil-mediated phagocytic activity
   compared with non-protective antibodies, highlighting the importance of
   sGP in vaccine protection and monoclonal antibody therapeutics against
   Ebola virus [[145]46]. Therefore, we and others pointed to a potential
   role of vaccine-induced, antibody-mediated neutrophil phagocytosis, as
   a surrogate of protection against COVID-19.

   As the first line of defense against pathogen, neutrophils play an
   important role during the acute phase of infection. To analyse the
   potential neutrophil activation during the acute phase of SARS-CoV-2
   infection, comparative transcriptional analyses were performed using
   PBMC samples of infected individuals within five days upon disease
   onset versus uninfected individuals. Consistent with previous
   single-cell RNA-seq analysis, PBMC transcripts from infected
   individuals revealed remarkably elevated expression of genes related
   neutrophils activation, neutrophil degranulation, and neutrophil
   extracellular trap formation, along with increased FcγR-mediated
   phagocytosis, chemokine signalling and BCR signalling pathways,
   compared to non-infected individuals. The immune signatures associated
   with neutrophils and FcγR-mediated phagocytosis were also upregulated
   upon infection. Therefore, our transcriptional data further support
   that ADNP is a potential significant biomarker for disease prevention
   and control of COVID-19.

   We also observed that non-BTI group had a higher level of ADCP
   responses than that in BTI group, suggesting that ADCP might also play
   an important role in preventing infection. ADCP exerts a pivotal
   antibody-mediated immune mechanism, where phagocytic cells (e.g.
   macrophage, neutrophils, dendritic cells) actively engulf and destroy
   target cells or pathogens that are opsonized by antibodies. Previous
   studies have shown that the RBD-specific antibody-dependent monocyte
   phagocytosis was lower in deceased patients compared to survived
   patients [[146]21,[147]47], which is consistent with our observations.
   It was found that the uptake of SARS-CoV-2 by macrophage via ADCP did
   not lead to an aberrant cytokine production, possibly due to the small
   size of the virus-IgG immune complex [[148]46]. Furthermore, neutrophil
   and macrophage depletion resulted in the loss of protection from the
   convalescent plasma. It was observed that Fc-mediated effector
   functions of plasma with weak neutralizing activities could protect
   mice from lethality and severe disease better than a control treatment
   of plasma with weaker Fc activities [[149]47]. Furthermore, neutrophil
   and macrophage depletion resulted in the loss of protection from the
   convalescent plasma, which both validating the essential role of these
   innate effector cells via ADCC and ADCP protection [[150]39].

   There are some limitations to our study. Firstly, in our study, our
   established predictive models were validated only among vaccine
   recipients who received inactivated COVID-19 vaccines and experienced
   breakthrough infections with Omicron BA.5 variants. It is important to
   further validate the protective role of vaccine-induced ADNP responses
   across different vaccine platforms and emerging variants. Secondly,
   antibody isotypes, glycosylation patterns, and FcR subtypes on multiple
   Fc effector functions could also regulate the neutrophil phagocytosis
   and confer disease attenuation, which were not dissected in this study.

   In summary, our study identified Fc-mediated neutrophil phagocytosis as
   a significant immunological correlate that contributes to disease
   attenuation in the absence of neutralization, which may provide key
   insights to guide effective pan-variant SARS-CoV-2 vaccine design.

Supplementary Material

   Supplementary materials1116.docx
   [151]TEMI_A_2434567_SM5618.docx^ (378.5KB, docx)

Funding Statement

   This work was supported by the National Key Research and development
   program [2023YFC2309100]; National Natural Science Foundation of China
   [92269118, 92269205, 92369117]; Nanjing Drum Tower Hospital Outstanding
   Youth Cultivation Fund [2024-JCYJ-YP-01]; Scientific Research Project
   of Jiangsu Health Commission [M2022013]; Project of Chinese Hospital
   Reform and Development Institute, Nanjing University, Aid project of
   Nanjing Drum Tower Hospital Health, Education &Research Foundation
   [NDYG2022003], and Postgraduate Research & Practice Innovation Program
   of Jiangsu Province [021093002704, JX22014156].

Disclosure statement

   No potential conflict of interest was reported by the author(s).

References