TNF inhibitors affect the induction and maintenance of spike-specific B-cell responses after mRNA vaccination

Abstract

Objectives

Tumour necrosis factor inhibitors (TNFi) are widely used and effective as treatment for immune-mediated inflammatory diseases (IMIDs). However, TNFi therapy causes a faster waning of antibody responses following vaccination. The underlying cause by which TNFi affect humoral immunity remains to be elucidated. The formation of long-lasting, high-affinity antibodies after vaccination results from germinal centre (GC)-derived, T cell-dependent B-cell responses. Therefore, this study investigated how TNFi affect the formation and maintenance of antigen-specific B- and CD4+ T-cell responses following SARS-CoV-2 mRNA vaccination.

Methods

SARS-CoV-2 spike-specific B-cell responses were characterised using spectral flow cytometry. Spike-specific CD4+ T cells were measured using an activation-induced marker assay. 15 patients with inflammatory bowel disease (IBD) treated with TNFi were compared with 9 IBD patients without systemic immunosuppression and 10 healthy controls.

Results

Spike-specific CD4+T-cell frequency and phenotype, including T follicular helper cells, were not affected by TNFi. Total spike-specific B-cell frequencies were reduced in TNFi-treated patients. Deep phenotyping revealed lower IgG+memory B-cell frequencies in TNFi-treated patients 3–6 months after vaccination. These data were confirmed in TNFi-treated rheumatoid arthritis patients. Interestingly, already at day 7 after the second vaccination, TNFi therapy reduced the induction of class-switched CD11c- CD71+activated B cells, which are believed to be GC-derived. Conversely, CD11c+B cells, associated with extrafollicular B-cell responses, were not affected by TNFi therapy.

Conclusions

These data suggest that TNFi therapy affects the differentiation of GC-derived B cells, which may explain its effect on humoral immune responses.

WHAT IS ALREADY KNOWN ON THIS TOPIC.

WHAT THIS STUDY ADDS

• Inflammatory bowel disease and rheumatoid arthritis patients treated with TNF inhibitors have lower antigen-specific IgG+ memory B-cell frequencies 3–6 months following SARS-CoV-2 vaccination.

• Treatment with TNF inhibitors reduces germinal centre-derived CD11c- CD71+ activated B-cell frequencies, whereas CD11c+B-cell subsets remained unaffected at day 7 after vaccination.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• These findings provide deeper insights into how treatment with TNF inhibitors in immune-mediated inflammatory disease (IMID) patients results in the rapid waning of antibodies after antigen exposure.

• Vaccine responses in IMID patients treated with TNF inhibitors should be monitored closely, and additional booster vaccinations might be necessary.

Introduction

Tumour necrosis factor inhibitors (TNFi) are commonly used in the treatment of several immune-mediated inflammatory diseases (IMIDs), such as inflammatory bowel disease (IBD), rheumatoid arthritis (RA), psoriasis, ankylosing spondylitis and psoriatic arthritis. They are often very successful in alleviating clinical symptoms of IMIDs. TNFi are available as monoclonal antibodies (infliximab, adalimumab, golimumab and certolizumab pegol) or as a soluble receptor protein (etanercept). Mechanistically, they all block interactions of soluble TNF-α with the cellular TNF receptors 1 and 2 (TNFR1 and TNFR2, respectively).¹ TNF is a pleiotropic cytokine crucial in many immunological processes, including activation, signalling and migration of lymphocytes.² However, the exact biological mechanisms by which TNF inhibition affects immune responses in IMIDs remain unclear.

An elegant way to investigate the effect of TNFi on the human adaptive immune response is by studying the response to vaccination. Previous studies using vaccines against pneumococcal disease, influenza or hepatitis A or B virus have suggested that TNFi treatment leads to lower humoral vaccine responses.³,⁷ The COVID-19 pandemic and resulting vaccination programmes provided a unique opportunity to investigate the effect of TNFi treatment to a de novo antigen on a large scale. Numerous studies on SARS-CoV-2 vaccination indicated that TNFi treatment leads to lower antibody induction and maintenance over time.⁸,¹¹ Furthermore, repeated SARS-CoV-2 vaccination resulted in lower IgG4 skewing in TNFi-treated patients compared with controls.¹²

The production of circulating antibodies by antibody-secreting cells (ASCs) provides protection against pathogens.^{13 14} ASCs are generated when antigen-specific B cells are activated by antigen and subsequently receive T-cell help from follicular helper cells (Tfh). B cells can differentiate into short-lived ASCs via an extrafollicular response to secure rapid, but less high-affinity, antibody formation. Additionally, germinal centres (GCs) are formed in secondary lymphoid organs, such as spleen and lymph nodes. The GC reactions ensure the production of high-affinity antibodies through somatic hypermutation and antigen- and Tfh-driven selection of B cells with high-affinity B-cell receptors (BCRs).¹⁵,¹⁷ In addition to ASCs, memory B cells (MBCs) are formed and can rapidly reactivate on reinfection, helping prevent (severe) disease. To understand how TNF inhibition affects antibody responses, recent research has focused on antigen-specific B-cell responses in patients treated with TNFi. They showed that TNFi induce normal antigen-specific B-cell frequencies within 1 month after vaccination, whereas long-lasting MBC formation may be hampered.^{8 18 19} However, these studies focused on total B cells or classical (IgD- CD27+) MBCs, thus not capturing the complexity of the antigen-specific B-cell response.

The different steps in B-cell differentiation upon antigen encounter remain in part to be elucidated, but recent progress has been made. Activated B cells (ActBCs) that arise after antigen encounter were recently identified as a population of CD27+CD71+ CD21 lo B cells and are presumed to precede GC-derived MBC and ASC formation.²⁰,²² ActBCs peak in circulation shortly after antigen exposure and contract again over time.^{2021 23},²⁶ A population of CD11c+CD21-/lo B cells is thought to form mainly via the extrafollicular route and was first described in the context of aberrant immune responses, such as during chronic infections and autoimmunity.^{27 28} However, increasing evidence suggests that CD11c+B cells are also part of normal immune responses.²¹²⁹,³¹

To better understand the underlying mechanism through which TNFi affect antibody responses, deep immune profiling of SARS-CoV-2 spike-specific B and CD4+T cells over time was performed following two SARS-CoV-2 vaccinations in IBD patients treated with adalimumab or infliximab. IBD patients without systemic immunosuppressive treatment and healthy donors were used as controls. Additionally, since TNFi are frequently prescribed in combination with other immunosuppressants, our findings were validated in IBD patients receiving TNFi with purine antagonists (PA) and in RA patients treated with TNFi and methotrexate (MTX).

Methods

Study design and patient inclusion

This study is a substudy of the Target-to-B! (T2B), a prospective multi-centre cohort study on SARS-CoV-2 vaccination in patients with various immune-mediated inflammatory diseases.¹⁰ This study was approved by the medical ethical committee (NL74974.018.20 and EudraCT 2021-001102-30, local METC number: 2020_194) and registered at the Dutch Trial Register (Trial ID: NL8900). All participants provided written informed consent. IBD patients were recruited at the Academic Medical Centre Amsterdam. Patients were diagnosed by a certified gastroenterologist and were treated with adalimumab or infliximab at least 6 months before their first SARS-CoV-2 vaccination. IBD patients without systemic immunosuppressive therapy, including systemic corticosteroids but excluding aminosalicylates, were included as IBD controls. Additionally, IBD patients treated with PA monotherapy or TNFi combination therapy (TNFi+PA) were included, as well as RA patients with MTX monotherapy, or TNFi combination therapy (TNFi+MTX). Age- and sex-matched healthy controls (HCs) were also recruited. The exclusion criteria were as follows: concomitant treatment with other systemic immunosuppressants, pregnancy and previous SARS-CoV-2 infection (self-reported positive PCR and/or positive anti-receptor binding domain (RBD) antibodies before the first vaccination and/or positive anti-nucleocapsid protein antibodies). If participants changed immunosuppressive treatment or became infected with SARS-CoV-2 during the study, they were excluded from further analysis. Participants were vaccinated between April 2021 and June 2021 with the mRNA-1273 (Moderna) vaccine at a 6-week interval. Peripheral blood samples were collected by venipuncture 7–12 days and 3–6 months after the second vaccination. Peripheral blood mononuclear cells (PBMCs) were isolated within 12 hours and frozen in liquid nitrogen until use.

Patient and public involvement

At what stage in the research process were patients/the public first involved in the research and how?

Patients and patient groups have been involved throughout the study. The T2B consortium is a partnership between researchers, clinicians and various patient groups in which patients’ representatives are involved in a dedicated work package. This contributes to new research ideas, determines priorities and guarantees maximal outreach of our results to the community. During the study, participants were notified about their SARS-CoV-2 antibody results. The study results were also shared during multiple patient-centred update meetings and by mailing to patient groups, who further disseminated our results. Our data were shared with the Dutch national health institute and greatly shaped the national vaccine campaign for IMID patients.

How were the research question(s) and outcome measures developed and informed by their priorities, experience and preferences?

To determine vaccine efficacy, in addition to the humoral and cellular vaccine responses in IMID patients, the T2B-COVID study also focused on the concerns of patients that vaccines might worsen their underlying diseases. To this end, we introduced various disease-specific patient-reported outcome measures to assess the activity of the underlying disease after vaccination and COVID-19.

How were patients/the public involved in the design of this study?

As described above, patients were directly involved in both conceptualisation of the T2B-COVID study, as well as members of the review committee of the National Grand Agency (ZonMw).

How were they involved in the recruitment to and conduct of the study?

Patients were involved in various ways. For example, patients at Reade Centre for Rheumatology research were asked to introduce an age- and sex-matched acquaintance to participate as an HC in this study. This led to active participation from both patients and HCs who were invested in the proper conduct of the study.

Were they asked to assess the burden of the intervention and time required to participate in the research?

The burden of participation was limited to a few time points of blood sampling (many of these combined with vaccination time points) and filling in online questionnaires.

How were (or will) they be involved in your plans to disseminate the study results to participants and relevant wider patient communities?

As stated above, our dissemination strategy included the sharing of individual results with participants as soon as the results are known. In addition, patient groups were separately informed on our main study findings, and we organised seminars for patients to disseminate our results.

Anti-RBD ELISA

Three different in-house developed ELISA assays were performed to detect SARS-CoV-2-specific IgG antibodies, as has been described before.^{32 33} In short, a quantitative ELISA was conducted to measure anti-RBD IgG levels. The serum samples from our participants were compared with a serially diluted calibrator consisting of pooled convalescent plasma, and ELISA results were expressed in arbitrary units per millilitre. A semi-quantitative total antibody bridging ELISA was used to detect individuals with SARS-CoV-2 infection before their first vaccination. This assay demonstrated increased sensitivity in very low antibody ranges (98.1% sensitivity and 99.5% specificity) compared with the anti-RBD IgG ELISA. A semi-quantitative total antibody bridging ELISA against nucleocapsid antibodies was used to detect SARS-CoV-2 infection during the follow-up of our study.

Activation-induced marker assay

PBMCs were thawed in Iscove’s Modified Dulbecco's Medium (IMDM) with 5% foetal calf serum (FCS), 1% penicillin/streptavidin and 0.1% DNAse, after which the cells were centrifuged and resuspended in IMDM with 5% FCS, 5% human serum and 1% penicillin/streptavidin. PBMCs were plated at 2×10∧6 cells per well and stimulated with a spike protein peptide pool (15-mer with 11aa overlap, Miltenyi Biotec; at 0.6 nmol of each peptide/mL) for 24 hours at 37°C. Anti-CD40 antibody (1 µg/mL, Miltenyi Biotec) was added for 15 min, just before peptide stimulation. Dimethyl sulfoxide (DMSO) and anti-clusters of differentiation 3/8 (CD3/CD28; Thermo Fisher) stimulation were used as negative and positive controls, respectively. After stimulation, PBMCs were stained with a panel consisting of 18 fluorescent conjugated antibodies and one viability dye (online supplemental table 2) for 30 min at room temperature (RT) in the dark. Cells were washed and fixated (Foxp3 staining buffer set, eBioscience) for 30 min at 4°C. Subsequently, PBMCs were washed and acquired on a FACSymphony (BD). Rainbow beads (BD) were used as reference to correct voltage settings during each acquisition. FlowAI R-plugin was used for data quality control; this included anomaly detection via flow rate check (timestep 1/10 s), signal acquisition (threshold, 1300) and dynamic range (upper and lower limits). Data were manually analysed using FlowJo v10 (FlowJo). DMSO background was subtracted from S-induced CD4+T-cell percentages during analysis. To capture as many spike-specific CD4+T cells as possible, five activation markers were used into one AIM output via Boolean OR gating: double-positive cells from all 10 possible dual AIM combinations were added up, whereas redundant cells were removed.

B-cell spectral flow cytometry staining

Antigen probe design, purification and antigen-specific B-cell staining were performed as previously described.^{22 25 30 34 35} Briefly, PBMCs were thawed in IMDM (Lonza) containing 10% FCS (Bodinco BV). Then, 10×10⁶ PBMCs were depleted of CD3+cells using EasySep Human CD3 Positive Selection Kit II (StemCell Technologies) as per manufacturer’s instructions. To stain antigen-specific B cells, biotinylated protein antigens SARS-CoV-2 spike-2P (prefusion stabilised trimer), SARS-CoV-2 RBD, nucleocapsid protein, influenza haemagglutinin (H1N1pdm2009), prefusion stabilised glycoprotein from respiratory syncytial virus (DS-Cav1) and tetanus toxoid (Vcar-Lsx003) were individually multimerised with fluorochrome-conjugated streptavidin in a 2:1 molar ratio at 4°C for 1 hour on a shaker. Then, 10% D-biotin (GeneCopoeia) was added to each multimerised protein antigen and incubated at 4°C for 30 min to minimise cross-reactivity. Biotinylated antigens were conjugated to two fluorochromes, except RBD, which was conjugated to a single fluorochrome (online supplemental table 3). A 31-colour spectral cytometry panel, including the six biotinylated antigens, was designed to immunophenotype antigen-specific B cells (online supplemental table 4). Samples were first stained with Live/Dead Fixable Blue Stain Kit (Invitrogen) in phosphate-buffered saline for 30 min at 4°C. Cells were washed with staining buffer containing 1% bovine serum albumin and 1 mM EDTA in phosphate-buffered saline and then stained with the above-mentioned spectral cytometry panel, including the protein antigens, for 30 min at 4°C. Cells were washed with staining buffer, fixed with cold paraformaldehyde 1% for 10 min at RT, and washed again twice with staining buffer. Data were acquired on Cytek Aurora 5 L using SpectroFlo software (Cytek Biosciences).

Computational flow cytometry: B-cell data pre-processing

Following spectral unmixing (SpectroFlo v3.0.1, Cytek), Flow cytometry files were loaded into OMIQ software from Dotmatics. Data were transformed using arcinh with cofactors adequate for each parameter to display a Gaussian distribution of negative populations while maximising the range of positive values. Initial gating was performed to select single, live, CD19+ cells, and to removedump and autofluorescent negative cells (dump markers included CD3, CD4, CD14 and CD56) (online supplemental figure 3A). PeacoQC was run to detect and remove flow cytometry anomalies in both signal acquisition and dynamic range.³⁶ To exclude batch effects, all data were normalised using Cytonorm for eight markers (CD20, CD38, CD19, CD24, human leukocyte antigen–DR, CD 45 receptor B (CD45RB), IgD and IgM) based on a cryopreserved reference HC sample, which was thawed, stained and measured with each batch to control for signal variation.³⁷ Subsequently, antigen-specific B cells were gated based on their specific combination of fluorochrome-conjugated streptavidin and were negatively gated for all other fluorochrome combinations; RBD-positive cells were gated from spike-positive cells (online supplemental figure 3B). For downstream analysis, data were subsampled to include only antigen-specific B cells.

Computational flow cytometry: dimensionality reduction and FlowSOM clustering

To acquire sufficient input data for uniform manifold approximation and projection (UMAP) visualisation and FlowSOM clustering, antigen-specific B-cell data from 104 participants from the T2B study cohort were included.¹⁰ This cohort included IMID patients using different immunosuppressants (ISPs), such as MTX and TNFi, IMID patients without ISP and HCs. For UMAP visualisation and FlowSOM clustering (with dimensions xdim=16 and ydim=16), 12 lineage markers (CD20, CD21, CD27, CD138, CD38, CD24, CD45RB, CD11c, IgM, IgA, IgG and IgD) were given as input. Consensus meta clustering was run to generate k=50 meta clusters, which were manually merged into 38 clusters and 16 main populations based on biological relevance. The clusters were annotated based on UMAP visualisation and heatmap analysis of the median marker expression between the clusters (annotations are listed in online supplemental table 5). Further downstream analyses on spike- and RBD-specific B cells were performed using only data from study participants included in this study.

Statistical analysis

Statistical analyses were performed using GraphPad Prism version 6.04 for Windows. Statistical significance was determined using the Wilcoxon ranked sum test for unpaired comparisons and the Wilcoxon signed rank test for paired comparisons. Dunn’s correction for multiple comparisons was used to adjust for multiple testing. P-values <0.05 were considered statistically significant.

Results

TNFi affect antibody maintenance

In this study, 15 IBD patients treated with TNFi monotherapy, of which three with infliximab and 12 with adalimumab, were included, as well as age- and sex-matched IBD patients without systemic immunosuppressive treatment (IBD controls; n=9), and HCs (n=10). The characteristics of the study participants are listed in table 1. Patients were stable on treatment or, in the case of IBD controls, stable on the absence of systemic immunosuppressive treatment. If changes in medication were made during the course of the study, participants were excluded from further analyses. Of the included IBD patients, most TNFi-treated patients were diagnosed with Crohn’s disease (CD; 12/15), whereas most of the IBD controls had ulcerative colitis (UC; 7/9). All TNFi-treated patients had detectable infliximab or adalimumab serum titres on the day of the first vaccination (V1pre) (table 1). All participants received two SARS-CoV-2 mRNA-1273 vaccinations at a 6-week interval. PBMCs were collected at baseline (V1pre), 7 days after the second vaccination (V2D7) and at 3 (V2M3) or 6 months (V2M6) after the second vaccination (figure 1A). Serum samples were collected at additional time points to measure antibody titres (figure 1A). IBD did not affect the antibody response as anti-RBD IgG titres were similar between HCs and IBD controls (figure 1B, C). Conversely, TNFi-treated patients displayed significantly reduced anti-RBD IgG titres 28 days and 6 months after the second vaccination, suggesting an aberration in antibody induction and maintenance in this group. To further emphasise the inhibitory effect of TNFi treatment on antibody maintenance following vaccination, TNFi combination treatment groups were added to our analysis, including IBD patients treated with either PA (n=15) or TNFi+PA (n=6), and RA patients treated with MTX (n=16) or TNFi+MTX(n=13) (online supplemental table 1). Comparison of SARS-CoV-2 antibody titres between these groups revealed that TNFi in combination with PA or MTX led to reduced antibody titres at V2M6, but not at V2D28, when compared with their respective PA or MTX monotherapy control group (online supplemental figure 1A,B). Taken together, these data indicate that TNFi treatment suppresses antibody responses following vaccination.

Table 1. Participant characteristics.

	IBD patients		Healthy control (n=10)
	TNF inhibitor (n=15)	IBD control (n=9)
Age in years, mean (SD)	41 (15)	45 (9)	44 (11)
Female sex, n (%)	5 (33%)	3 (33%)	2 (20%)
Disease, n (%)
Ulcerative colitis	3 (20%)	7 (70%)	–
Crohn’s disease	12 (80%)	3 (30%)	–
TNF inhibitor type, n (%)
Infliximab	3 (20%)	–	–
Adalimumab	12 (80%)	–	–
TNF inhibitor treatment duration in months, median (IQR)	26 (17–42)	–	–
TNF inhibitor titre at baseline (V1pre) in µg/mL, mean (SD)
Infliximab	22.2 (11)	–	–
Adalimumab	11.7 (5.6)	–	–
Days between the second vaccination and V2D7 PBMC collection, median (IQR)	7 (7-8)	7 (7-11)	7 (7-7)
Days between the second vaccination and V2M3-6 PBMC collection, median (IQR)	196 (189–200)	199 (194–208)	185 (183–191)
Seroconversion after the second vaccination, n (%)	14 (100%)	10 (100%)	11 (100%)
Anti-RBD IgG titre at V2D28, median (IQR)	160 (120–204)	269 (250–303)	234 (190–297)
Anti-RBD IgG titre at V2M6, median (IQR)	4.5 (3.9–8.3)	32 (28–36)	30 (20–41)

IQR, interquartile range; PMBCs, peripheral blood mononuclear cells; TNF, tumour necrosis factor; V2D7, 7 days after the second vaccination; V2D28, 28 days after the second vaccination; V2M6, 6 months after the second vaccination; V2M3-6, 3 or 6 months after the second vaccination; V1pre, baseline.

Figure 1. SARS-CoV-2 antibody responses. (A) Schematic overview of the study design. PBMCs were collected at baseline (V1pre), 7 days after the second SARS-CoV-2 mRNA-1273 vaccination (V2D7) and 3 or 6 months after the second vaccination (V2M3 or V2M6). (B) Anti-RBD antibody titre over time. (C) Anti-RBD antibody titres per timepoint. The dotted line represents seroconversion level. The horizontal bars represent the median. PBMCs, peripheral blood mononuclear cells; RBD, receptor binding domain.

TNFi do not affect vaccine-specific CD4+ T-cell induction

As CD4+T cells are important for the generation of high-affinity antibodies, we analysed the spike-specific CD4+T-cell response using an AIM assay (figure 2A). PBMCs were stimulated overnight with a 15-mer peptide pool covering the complete spike protein, and CD4+T cells positive for at least two AIMs were defined as spike-specific (figure 2A). At V2D7, the frequency of spike-induced AIM+CD4+ T cells was similar between TNFi-treated IBD patients and IBD controls (figure 2B). Given that TNFi may potentially affect spike-specific CD4+T-cell phenotype while total frequencies remain unaffected, the spike-specific CD4+Th distribution was then investigated based on chemokine expression, including circulating Tfh (cTfh; CXCR5+), Th1 (CXCR5-CXCR3+CXCR6-), Th2 (CXCR5-CXCR3-CXCR6-CCR4+), Th17 (CXCR5-CXCR3-CXCR6+CCR4+), Th9 (CXCR5-CXCR3-CXCR6+CCR4-) and Th1-like (CXCR5-CXCR3+CXCR6+) cells (figure 2C). Th17 cells were most abundantly present within our AIM+CD4+ T-cell population, followed by cTfh, Th1-like and Th1 (figure 2D). Frequencies of the different Th subsets did not differ between the TNFi-treated and IBD control groups (figure 2E), also not when stratified for age, IBD type (CD vs UC) or TNFi type (adalimumab vs infliximab) (data not shown). Moreover, PD1 and ICOS co-expression on total S-induced AIM+CD4+T cells and AIM+CXCR5+ CD4+cTfh cells, which is indicative of CD4+T(f)h activation, did not differ between TNF-treated patients and IBD controls (figure 2F and G). Lastly, spike peptide pool-induced TNF-α, TNF-γ and IL-2 co-expression in total CD4+T cells was measured as another proxy for spike-specific CD4+T-cell activation. Similar to our AIM+CD4+ T-cell frequencies, CD4+T-cell expression of TNF-α, interferon-γ and interleukin (IL)-2 was similar between TNFi-treated patients and IBD controls (figure 2H). Collectively, these data conclude that IBD patients treated with TNFi have similar antigen-specific CD4+T-cell induction and Th phenotype following vaccination, as compared with untreated IBD controls.

Figure 2. SARS-CoV-2 CD4+T-cell induction is not affected by TNF inhibitor treatment. (A) Schematic design of the activation-induced marker (AIM) assay, combining 10 different dual-AIM combinations into one total AIM. (B) Percentage of total S-induced AIM+CD4+ T cells. The background was subtracted. (C) Representative gating strategy to identify S-induced AIM+T follicular helper, Th1, Th1-like, Th9, Th17 and Th2 cells. (D) Distribution of the Th phenotype within S-induced AIM+CD4+ T cells. (E) Comparison of the different AIM+Th subsets between TNFi-treated patients and IBD controls. (F) Representative flow cytometry plots of PD1 and ICOS expression on AIM+CD4+ T cells and AIM+Tfh cells at V2D7. (G) Percentage of PD1 and ICOS expressing AIM+CD4+ T cells and AIM+Tfh cells at V2D7. (H) Percentage of total CD4+T cells expressing two or three of the following cytokines: IL-2, TNF-α and IFN-γ. Plotted is the median with IQR. AIM, activation-induced marker; IBD, inflammatory bowel disease; ICOS, inducible T-cell costimulatory; IFN-γ, interferon gamma; IL-2, interleukin-2; IQR, interquartile range; PD1, programmed cell death protein 1; Th, T helper cell; Th1, T helper cell; Th17, T helper 17 cell; Th2, T helper 2 cell; Th9, T helper 9 cell; TNFi, tumour necrosis factor inhibitors; TNF-α, tumour necrosis factor alpha.

TNFi treatment results in lower spike-specific memory B cells

SARS-CoV-2-specific antibody maintenance was significantly affected by TNFi treatment. Therefore, frequencies and dynamics of the spike-specific B-cell compartment following SARS-CoV-2 vaccination were investigated. To enable deep phenotyping of antigen-specific B cells, PBMCs were stained with a 31-marker flow cytometry panel including dual-labelled spike- and RBD-specific probes (online supplemental figure 3A,B, online supplemental tables 3,4). Seven days following the second vaccination, TNFi-treated patients had lower RBD-specific B-cell frequencies than IBD controls, whereas spike-specific B cells showed a lower, but not statistically significant, trend (figure 3A). Strikingly, 3–6 months later, TNFi-treated patients displayed significantly lower frequencies of total RBD- and spike-specific B cells (figure 3B, online supplemental figure 3C). Spike- and RBD-specific B-cell frequencies did not differ between HCs and IBD controls at both V2D7 and V2M3-6. Similar effects were observed in patients treated with TNFi in combination with MTX or PA (online supplemental figure 3D,E).

Figure 3. SARS-CoV-2 B-cell frequencies are affected by TNFi treatment (A) Spike- and RBD-specific B-cell frequencies at 7 days and (B) 3–6 months after the second vaccination (V2M3-6). (C) UMAP representation of all antigen-specific B cells. (D) UMAP representation of antigen-specific B cells per timepoint. (E) Stacked bar graph representing the average proportion of the 16 S+B-cell clusters as a percentage of total CD19+B cells. (F–H) Frequency of spike-specific IgG+memory B cells from total spike-specific CD19+B cells compared between (F) TNFi and IBD controls at V2M6, (G) TNFi+MTX and MTX monotherapy at V2M6, and (H) TNFi and purine antagonists (PAs) at V2M3. Plotted is the median with IQR. IQR, interquartile range; MTX, methotrexate; RBD, receptor binding domain; TNFi, tumour necrosis factor inhibitor; UMAP, uniform manifold approximation and projection; V2M3, 3 months after the second vaccination; V2M6, 6 months after the second vaccination.

Next, deep phenotyping of the spike-specific B-cell compartment was performed to uncover which B-cell populations were affected by TNFi treatment. First, UMAP was employed to visualise the spike-specific B-cell compartment (figure 3C). Subsequently, FlowSOM clustering was performed using a selection of B-cell markers (CD20, CD21, CD27, CD138, CD38, CD24, CD45RB, CD11c, IgM, IgA, IgG and IgD), which revealed 38 clusters that were manually merged into 16 main B-cell populations based on surface marker expressions (online supplemental figure 3F).^{25 38} UMAP representation of the 16 populations showed a clear distribution of naïve(-like) B cells (IgD⁺CD27^-CD38^-CD21⁺), ASCs (CD27⁺CD38⁺CD20^lo), MBCs (CD27⁺CD38^-CD21⁺), ActBCs (CD27⁺CD38^loCD21^loCD71⁺) and double negative 2 B cells (DN2); CD27^-CD21^-IgD^-CD11c⁺) (figure 3C). This clustering demonstrated that spike-specific B-cell populations were highly dynamic over time: at V2D7, the majority of the spike-specific compartment comprised ActBCs, ASCs and DN2, whereas IgG+MBCs dominated the spike-specific response 3–6 months later (figure 3D and E).

Since we observed lower spike-specific B-cell frequencies in TNFi-treated patients 3–6 months after SARS-CoV-2 vaccination, we investigated the effect of TNFi on the spike-specific IgG+MBC response, which was the largest compartment at these timepoints. At 6 months, TNFi-treated patients displayed a trend of lower IgG+MBC frequencies than IBD controls (figure 3F). However, as only a small number of TNFi monotherapy patients were sampled at V2M6, we also studied IgG+MBC responses in TNFi+MTX-treated patients at V2M6. IgG+MBC frequencies were lower in MTX+TNFi-treated patients at V2M6 than those treated with MTX monotherapy (figure 3G). In addition, at V2M3, TNFi monotherapy patients had significantly lower IgG+MBC frequencies than PA monotherapy patients (figure 3H). IBD controls were not sampled at V2M3, but PA-treated patients displayed similar frequencies of total spike-specific B cells and spike-specific IgG+MBCs at V2M6 compared to IBD controls (online supplemental figure 3E-G). Similar to IgG+MBCs, IgA+MBC frequencies were significantly reduced in TNFi-treated patients when compared with PA monotherapy patients at V2M3, but not when compared with IBD controls at V2M6 (online supplemental figure 3H-J). Together, these data show that TNFi treatment results in lower IgG+MBC frequencies 3–6 months following SARS-CoV-2 vaccination.

TNFi-treated patients exhibit lower antigen-specific CD11c- B-cell frequencies, whereas CD11c+ B-cell levels are maintained

As the lower IgG+MBC frequencies in TNFi-treated patients may be the result of less induction of long-lived MBC formation, the composition of the antigen-specific B-cell compartment early after vaccination was investigated. At V2D7, a significantly higher relative frequency of transitional B cells (CD27- CD38+) and IgD/IgM+ActBC (CD27+CD38 lo CD71+/-CD24±CD11c+/-) was observed in the TNFi-treated group (figure 4A–B). All other populations, including naïve B cells, IgG/IgA/IgM+MBC, IgG/IgA/IgM+ASC, IgG/IgA+DN2, DN3 and IgG/IgA+ActBC clusters were similarly expressed between TNFi-treated patients and IBD controls at V2D7 (figure 4A, online supplemental figure 4A-F).

Figure 4. Activated B cells are less induced by TNFi treatment, whereas CD11c+B cells are not. (A) Volcano plot displaying S+B-cell populations differentially expressed between TNFi and IBD controls 7 days after the second vaccination (V2D7). (B) Percentage of spike-specific transitional B cells and IgM/IgD+activated B cells at V2D7 between TNFi and IBD controls. (C) UMAP representation and clustering of antigen-specific IgG/IgA+activated B cells (ActBCs) and the overlay of CD11c expression projected on the UMAP. (D) Percentage of spike-specific IgG+CD11c- and CD11c+ActBCs at V2D7 compared between TNFi and IBD controls. (E) Percentage of spike-specific IgA+CD11c- and CD11c+ActBCs at V2D7 compared between TNFi and IBD controls. (F) Percentage of total spike-specific IgG+ActBCs at V2M6. (G) Percentage of CD11c- and CD11c+IgG+ ActBCs at V2M6. Plotted is the median with IQR. ActBCs, activated B cells; IBD, irritable bowel disease; IQR, interquartile range; TNFi, tumour necrosis factor inhibitor; UMAP, uniform manifold approximation and projection; V2D7, 2 days after the second vaccination; V2M6, 6 months after the second vaccination.

Given that ActBCs are a population of antigen-specific B cells potentially contributing to the pool of resting MBCs,²⁰,²² the effect of TNFi therapy on the ActBC compartment was investigated in more detail. ActBCs largely contributed to the spike-specific response 7 days after the second vaccination, as previously demonstrated by our work and others (figure 3D–E).^{24 25 39} In the present study, the ActBC compartment consisted of six IgG+ActBC clusters and two IgA+ActBC clusters, including multiple CD11c-expressing populations (figure 4C, online supplemental figure 3F). Previously, in HCs, we demonstrated that the different ActBC clusters display distinct dynamics.²⁵ After the initial ActBC induction following vaccination, some clusters contracted 6 months later, whereas others did not.²⁵ The CD11c+ActBC clusters contracted significantly over time. Remarkably, TNFi therapy had differential effects on distinct ActBC clusters. TNFi-treated patients exhibited lower CD11c- ActBCs but similar CD11c+ActBCs frequencies compared with IBD controls (figure 4D). Besides CD11c+ActBCs, CD11c+IgG+ DN2 frequencies also remained unaffected by TNFi treatment, supporting the idea that TNFi treatment affects CD11c- IgG+ but not CD11c+IgG+ B-cell frequencies (online supplemental figure 4D). This observation was corroborated by the analysis of the IgA+ActBC response, where CD11c- IgA+ActBCs frequencies were also lower in the TNFi-treated group, whereas CD11c+IgA+ ActBC frequencies were similar between TNFi-treated patients and IBD controls (figure 4E). Likewise, CD11c+IgA+ DN2 B cells were not affected by TNFi treatment (online supplemental figure 4D). Lastly, although ActBCs are mainly present in the circulation shortly after antigen exposure, we could still detect spike-specific IgG+ActBCs 6 months after vaccination (figure 3D and 4F). Similar to V2D7, at V2M6, CD11c- IgG+ActBC frequencies were lower in TNFi-treated patients than in IBD controls (figure 4G). Together, our data demonstrate that following SARS-CoV-2 mRNA vaccination, TNFi-treated patients have lower antigen-specific class-switched CD11c- B-cell frequencies than non-treated patient controls, whereas class-switched CD11c+B cells remain unaffected.

Discussion

The effect of TNFi treatment on protective immunity and vaccination has not been fully explored. Our previous study and those of others has shown that TNFi treatment results in lower antibody maintenance following SARS-CoV-2 vaccination.^{9 11 19} This led us to hypothesise that TNFi may affect the formation of long-lived ASCs, which mainly rely on T-cell dependent GC-driven responses.^{40 41} Therefore, in this study, we analysed the effect of TNFi therapy in IBD patients on antigen-specific B-cell and CD4+T-cell responses following SARS-CoV-2 mRNA vaccination. In line with previous findings, TNFi treatment does not affect vaccine-specific CD4+T-cell frequency and phenotype shortly after vaccination,^{18 42} but does affect the long-lived MBC compartment.^{8 19 43} Our data now show that, already in the induction phase of the antigen-specific B-cell response, TNF inhibition leads to reduced CD11c- CD71+ActBC frequencies, whereas CD11c+B cells remain unaffected.

Previous studies in mice have indicated that TNF signalling is important for proper GC functioning. Studies using TNF-α-knockout mice or mice treated with TNFi revealed that TNF signalling is important for the formation of follicular dendritic cell (FDC) networks and GC development.⁴⁴,⁴⁸ Furthermore, an elegant study in humans demonstrated that etanercept therapy leads to less FDC and GC staining in tonsils.⁴³ Long-lived ASCs and MBCs are mainly formed via T cell-dependent GC responses.^{40 49 50} In line with the suggestion that TNF inhibition negatively affects GC responses, spondyloarthritis patients treated with TNFi were shown to display reduced humoral responses following T cell-dependent hepatitis B vaccination, whereas T cell-independent vaccination responses against pneumococcal polysaccharide were less affected.⁵¹ Given the indications that TNF inhibition affects GC responses, this study investigated the influence of TNFi therapy on the occurrence and dynamics of distinct antigen-specific B-cell differentiation stages following SARS-CoV-2 mRNA vaccination. Our data demonstrate that TNFi treatment induces less IgG+CD11c- ActBCs at day 7 after the second vaccination, demonstrating that GC-coined B-cell differentiation responses are affected by TNF inhibition.

Strikingly, IgG+CD11c+ActBC and CD11c+DN2 B-cell frequencies were not affected by TNFi treatment. The position of CD11c+B cells in the extrafollicular or GC response is still a matter of debate. The prevailing view suggests that CD11c+B cells lacking CD27 expression arise from the extrafollicular pathway, although this consensus has not yet been established for CD27+CD11c+B cells. The comparison of the level of Ig somatic hypermutation (SHM) in CD11c+vs CD11c- B cells may provide an indication of whether CD11c+B cells are derived from GCs; however, SHM can also occur to some extent extrafollicularly.^{52 53} Song et al (2022) showed that splenic CD11c+B cells following viral infection in mice exhibited minimal Ig clonal overlap with GC B cells,⁵⁴ whereas another study using human tonsils demonstrated similar SHM frequencies between CD27+CD11c+B cells and CD27+MBCs0.⁵⁵ Besides the EF- and GC response being mutually exclusive, CD11c+CD27+ B cells may arise from CD27+CD11c- MBCs that gained CD11c expression on reactivation. This hypothesis is supported by recent data that CD27+B cells can gain CD11c+expression in vitro.⁵⁶ Our data showing that both IgG+CD11c+ActBC and IgG+CD11c+DN2 s are not affected by TNFi, whereas the IgG+CD11c- ActBC compartment is affected, strongly support the GC-derived nature of IgG+CD11c- ActBCs and point to a more non-GC nature for IgG+CD11c+B cells. Longitudinal studies using single-cell BCR clonal mapping and RNA sequencing of the different B-cell phenotypes are needed to further elucidate the differentiation pathways of CD11c+B cells and demonstrate whether or not they can be derived from GCs.

Besides changes in the CD11c- ActBC and MBC compartment, we also observed relative increases in antigen-specific naïve-like and transitional B-cell frequencies in TNFi-treated patients following vaccination. This observation is in line with data from Anolik and coworkers (2008), who saw similar effects on total B-cell subsets in the peripheral blood of etanercept-treated RA patients.⁴³ One hypothesis for this observation could be that transitional and naïve B cells remain in the circulation because fewer GCs are formed. A similar phenomenon has been observed in an asplenic paediatric patient, in whom increased frequencies of transitional B cells were observed, likely due to lower niche availability.⁵⁷

The lower maintenance of antigen-specific MBC frequencies combined with the fast decline in antibody titres following vaccination in patients under TNFi therapy may imply that these TNFi-treated patients become more susceptible to disease on re-infection or breakthrough infection. Moreover, persisting GC reactions support the maintenance and clonal evolution of the MBC pool in response to emerging viral variants.^{58 59} Indeed, IMID patients treated with TNFi have an increased risk of COVID-19 breakthrough infection after completing the primary vaccination.^{60 61} However, TNFi treatment was also associated with decreased risk of severe COVID-19 when compared with other commonly used immunosuppressants, possibly due to the anti-inflammatory effects of TNF suppression on the systemic inflammatory response in COVID-19.^{62 63} Taking all of this into account, these patients may benefit from repeated booster vaccinations.¹⁰

We acknowledge several limitations in this study. While we included an untreated IBD control group, most of these controls had UC, whereas the majority of TNFi-treated patients had CD. Despite this difference, UC patients remain a suitable disease control group, as both conditions share fundamental immunological and inflammatory mechanisms.⁶⁴ Additionally, we were only able to include a limited number of longitudinal samples from TNFi patients, highlighting the need for further research to substantiate our findings.

In conclusion, TNFi treatment is associated with reduced antigen-specific B-cell induction and maintenance following mRNA vaccination. More specifically, spike-specific CD11c- ActBCs (CD71+CD21 CD27+) and resting classical MBCs are lower in TNFi-treated patients on mRNA vaccination, whereas CD11c+B cells remain unaffected. We therefore speculate that TNF signalling inhibition restricts GC development or functioning, whereas the short-lived extrafollicular B-cell response remains intact. Given that spike-specific cTfh frequencies remained unaltered in TNFi-treated patients, we hypothesise that the lower output of GC-derived B-cell responses is not due to a reduction in Tfh cells. With this knowledge, vaccine responses in these patients should be monitored closely, and additional booster vaccinations might be necessary.

Data availability statement

Data are available upon reasonable request.