Antibody decay, T cell immunity and breakthrough infections following two SARS-CoV-2 vaccine doses in inflammatory bowel disease patients treated with infliximab and vedolizumab

Anti tumour necrosis factor (anti-TNF) drugs increase the risk of serious respiratory infection and impair protective immunity following pneumococcal and influenza vaccination. Here we report SARS-CoV-2 vaccine-induced immune responses and breakthrough infections in patients with inflammatory bowel disease, who are treated either with the anti-TNF antibody, infliximab, or with vedolizumab targeting a gut-specific anti-integrin that does not impair systemic immunity. Geometric mean [SD] anti-S RBD antibody concentrations are lower and half-lives shorter in patients treated with infliximab than vedolizumab, following two doses of BNT162b2 (566.7 U/mL [6.2] vs 4555.3 U/mL [5.4], p <0.0001; 26.8 days [95% CI 26.2 – 27.5] vs 47.6 days [45.5 – 49.8], p <0.0001); similar results are also observed with ChAdOx1 nCoV-19 vaccination (184.7 U/mL [5.0] vs 784.0 U/mL [3.5], p <0.0001; 35.9 days [34.9 – 36.8] vs 58.0 days [55.0 – 61.3], p value < 0.0001). One fifth of patients fail to mount a T cell response in both treatment groups. Breakthrough SARS-CoV-2 infections are more frequent (5.8% (201/3441) vs 3.9% (66/1682), p = 0.0039) in patients treated with infliximab than vedolizumab, and the risk of breakthrough SARS-CoV-2 infection is predicted by peak anti-S RBD antibody concentration after two vaccine doses. Irrespective of the treatments, higher, more sustained antibody levels are observed in patients with a history of SARS-CoV-2 infection prior to vaccination. Our results thus suggest that adapted vaccination schedules may be required to induce immunity in at-risk, anti-TNF-treated patients.

V accination programmes have reduced SARS-CoV-2 transmission, hospitalisation and deaths 1 . Patients treated with immunosuppressive drugs were excluded from the original trials for COVID-19 vaccines 2,3 . Consequently, data relating to the magnitude and durability of immune responses and subsequent vaccine effectiveness in this population are limited 4 .
Drugs targeting tumour necrosis factor (TNF), such as infliximab, are the most frequently prescribed biologic therapies used in the treatment of immune-mediated inflammatory disorders (IMIDs). Observational studies indicate that most patients with inflammatory bowel disease (IBD), an archetypal IMID, mount serological responses following SARS-CoV-2 vaccines, although most were underpowered to discern the impact of specific drugs, including immunomodulators (azathioprine, mercaptopurine and methotrexate) and/or biologic therapies [5][6][7][8] . We reported that antibody responses following SARS-CoV-2 infection 9,10 or a single dose of either the BNT162b2 or ChAdOx1 nCoV-19SARS-CoV-2 vaccines were impaired in anti-TNF treated patients when compared to vedolizumab-treated patients 11 . Vedolizumab, is a gut-selective anti-integrin α4β7 monoclonal antibody that, unlike anti-TNF drugs, is not associated with increased susceptibility to systemic infection or attenuated serological responses to vaccination 12 .
In this work, we show that anti-SARS-CoV-2 spike antibody responses are attenuated and less durable following two doses of the BNT162b2 and ChAdOx1 nCoV-19 SARS-CoV-2 vaccines in infliximab-treated compared with vedolizumab-treated patients with IBD. Irrespective of biologic drug type, one-fifth of all patients do not mount a T cell response and a minority mount neither antibody nor T cell responses. Breakthrough SARS-CoV-2 infections, which are associated with lower antibody levels after the second dose of vaccine, are more common and occur earlier in infliximab-treated patients. Higher and more sustained antibody levels are observed in patients with a history of SARS-CoV-2 infection. Further work to define immunity after third primary and booster vaccine doses is needed to inform the need for adapted vaccination schedules in at-risk anti-TNF treated patients.

Results
Patient characteristics. Between September 22, 2020 and December 23, 2020, 7226 patients were recruited to the CLARITY study from 92 UK hospitals 10 . In this analysis we included 2279 infliximab-treated and 1031 vedolizumab-treated participants without a history of prior SARS-CoV-2 infection, who had received uninterrupted biologic therapy since recruitment and had an antibody test between 14 and 70 days after the second dose of either the BNT162b2 and ChAdOx1 nCoV-19 SARS-CoV-2 vaccines. Participant characteristics are shown in Table 1.
Additional analyses are presented for a subset of 211 infliximab-treated and 71 vedolizumab-treated patients included in our T cell experiments (Supplementary Table 1), and a further 530 infliximab-treated and 224 vedolizumab-treated participants who had a history of SARS-CoV-2 infection before vaccination (Supplementary Table 2).
After propensity matching for immunomodulator use and the other factors associated with choice of biologic, we confirmed lower anti-S RBD antibody concentrations in infliximab-treated compared to vedolizumab-treated patients (BNT162b2 600.1 U/ mL [6.0] (Fig. 2a, b respectively). Age ≥60 years and Crohn's disease were also independently associated with lower anti-S RBD antibody concentrations in vaccinated participants. Thiopurine or methotrexate use was independently associated with lower anti-S RBD antibody concentrations in participants who received the BNT162b2, but not the ChAdOx1 nCoV-19, vaccine. Current smoking, non-white ethnicity and steroid use were associated with lower anti-S RBD antibody concentrations in participants who received the ChAdOx1 nCoV-19 but not the BNT162b2 vaccine. To assess the effect of vaccine type on antibody responses, we combined our response data in a model that included vaccine type in addition to the significant factors above. Vaccination with the BNT162b2 vaccine compared to the ChAdOx1 nCoV-19 was independently associated with a 3.7 fold [95% CI 3.29-4.12] higher anti-S RBD antibody concentration (p < 0.0001) (Fig. 2c).
Seroconversion rates after the first vaccine dose were lower in infliximab-treated compared to vedolizumab-treated participants ( Fig. 1). However, administration of a second vaccine dose resulted in a >100-fold and >25-fold increase in antibody concentrations in recipients of the BNT162b2 and ChAdOx1 nCoV-19 vaccines, respectively ( Fig. 1). Overall, more infliximabtreated than vedolizumab-treated patients failed to seroconvert after their second vaccine dose (5.9% vs 1.3%, p < 0.0001). Seroconversion rates stratified by biologic therapy and vaccine type are reported in Supplementary Table 4.
Anti-spike T cell responses following two doses of BNT162b2 and ChAdOx1 nCoV-19 SARS-CoV-2 vaccines. There were no significant differences in the magnitude of anti-spike T cell responses observed in infliximab-treated compared with vedolizumab-treated patients after one or two doses of either vaccine (Fig. 3a). The proportion of patients failing to mount detectable T cell responses were similar in both groups (infliximab 19.6% vs. vedolizumab 19.2%). For recipients of one and two doses of the BNT162b2 vaccine, there was a modest positive correlation between T cell responses and antibody concentration. This association was not observed in recipients following either dose of the ChAdOx1 nCoV-19 vaccine (Fig. 3b). When T cell responses were ranked by magnitude of antibody responses, most patients who did not mount an antibody response after the first vaccine dose (indicated by the dark grey bar) had a detectable T cell response (Fig. 4). In addition to the uncoupling of the T cell and antibody responses demonstrated, this analysis emphasised that about one-fifth of participants made no T cell responses irrespective of vaccine used (indicated by the light grey bars). Moreover, a minority of individuals ( Overall, following two doses of either vaccine, anti-S RBD antibodies showed minimal decay to the last follow-up in patients treated with vedolizumab ( Fig. 5 and Supplementary Fig. 2) and were similar to those observed in participants in the Virus Watch community cohort ( Supplementary Fig. 3). However, in infliximab-treated participants, the geometric mean concentrations dropped to the seroconversion threshold by about 25 weeks after the second dose irrespective of the vaccine administered (Fig. 5). Infliximab compared to vedolizumab treatment, current smoking and white ethnicity were associated with a faster fall in anti-S RBD antibody concentration below the seroconversion threshold. (Supplementary Figs. 4,5).

Discussion
We have shown that in infliximab-treated patients, anti-SARS-CoV-2 spike antibody responses are attenuated following two doses of the BNT162b2 and ChAdOx1 nCoV-19SARS-CoV-2 vaccines. One-fifth of both infliximab-treated and vedolizumabtreated patients did not mount a T cell response and a small subset of patients had neither antibody nor T cell responses. Antibody half-lives were shorter in infliximab-treated patients. Breakthrough SARS-CoV-2 infections were more common and occurred earlier in infliximab-treated patients who received the ChAdOx1 nCoV-19 vaccine. The risk of breakthrough infection was predicted by lower antibody levels after the second dose of the vaccine. Irrespective of biologic treatment, higher and more sustained antibody levels were observed in patients with a history of SARS-CoV-2 infection.
Sustained antibody responses observed in vaccinated patients with a history of prior SARS-CoV-2 infection indicate that third antigen exposure enhances the serological response. This supports the rationale for prioritising a third dose of vaccine to clinically vulnerable patient populations [13][14][15][16] , who otherwise may face further periods of social distancing or hospitalisation following infection. Whilst drawing direct comparisons between IBD patients and patients treated with more potent chemotherapies is limited by the degree to which patients are immunosuppressed, data from solid organ transplant recipients shows that a third dose of vaccine also leads to sustained immune responses 17 .
Irrespective of biologic or immunosuppressant use, and in keeping with the original trials 2,18 , the highest antibody responses were seen in recipients of the BNT162b2 vaccine. Like in the general population, these responses waned more quickly than in the recipients of the ChAdOx1 nCoV-19 vaccine 19 . Unlike the general population 20 , but similar to renal transplant recipients 4 , we did not observe differences in T cell ELISpot responses between recipients of the BNT162b2 and ChAdOx1 nCoV-19 vaccines. The differences observed in breakthrough infection by vaccine type reported here are consistent with the differences in efficacy reported in the respective clinical trials 2,3,21 . The higher peak antibody levels and the lower rate of SARS-CoV-2 breakthrough infections suggest that the BNT162b2 rather than the ChAdOx1 nCoV-19 vaccine should be used for primary vaccination in infliximab-treated patients and, although untested, supports the use of BNT162b2 for third doses in all patients treated with an anti-TNF regardless of the primary vaccine type.
All patients treated with anti-TNF therapy should receive a third primary dose of the SARS-CoV-2 vaccine and our data support recent recommendations that this should occur about 4-8 weeks after the second dose 13,14,16 during periods of high transmission in the population. Our data demonstrate that patients treated with vedolizumab and infliximab-treated patients with prior SARS-CoV-2 infection have sustained antibody levels beyond 6 months.
When starting a biologic, it would be reasonable to consider differences in SARS-CoV-2 vaccine response as one of the factors when determining which drug to use. For patients who need to start anti-TNF therapy, the benefits of combination immunomodulator therapy should be weighed against the risk of attenuated vaccine response, and whenever feasible, patients should first receive a SARS-CoV-2 vaccine dose. Further research to determine whether timing third vaccine doses towards the end of anti-TNF treatment cycles when drug levels are lowest leads to greater immunogenicity 9 is needed. Other strategies including the temporary discontinuation of immunomodulators 22 , the use of heterologous vaccines 23 and adjuvants including the influenza vaccines (ComFluCOV) 24 need to be studied in immunosuppressed patient groups.
The biology underpinning loss of durable antibody responses and uncoupling of the B cell and T cell responses merit further research. TNF is a pleiotropic cytokine and its activities include maturation of antigen-presenting cells, modulation of T cell responses and stimulation of immunoglobulin synthesis [25][26][27] . TNF neutralisation, or genetic ablation, results in substantial loss of B-cells in primary follicles in germinal centres, reduced numbers of memory B-cells in the periphery but preserved numbers of T cells 25 . Uncoupling of humoral and T cell immunity to SARS-CoV-2 has been observed in healthy individuals 28 , and although the relative contributions of memory B cell and T cell responses have yet to be fully defined in SARS-CoV-2 immunity, the preservation of T cell immunity reported here should provide some reassurance for anti-TNF treated patients. However, it is noteworthy that one-fifth made no anti-spike T cell response following two doses of either vaccine. Chronic TNF exposure, a feature of many IMIDs, can render T cells anergic and can be reversed by anti-TNF treatment 29 . This may in part explain why the magnitude of T cell responses observed in anti-TNF-treated patients in this study did not differ significantly from patients treated with vedolizumab.
Although our data show major differences in the magnitude and durability of antibody responses, we have not assessed the impact of biologic therapy on specific immunoglobulin classes, antibody neutralisation or mucosal immune responses, which may be impaired, in particular, with anti-a4b7 therapy 30,31 . However, previous studies have demonstrated that anti-RBD antibody levels such as the ones measured in this study, strongly correlate with Wuhan Hu-1 live virus and variant S RBD neutralisation assays 32,33 , and we have demonstrated here that early antibody responses to vaccination correlates with the subsequent risk of breakthrough infection in immunosuppressed patients.
Infliximab was associated with attenuated, less durable vaccineinduced anti-SARS-CoV-2 spike antibody responses and a 50% increase in subsequent breakthrough SARS-CoV-2 infection. Further work to define immunity after third primary and booster vaccine doses is needed to inform the need for adapted vaccination schedules in at-risk anti-TNF treated patients.

Methods
Patient and settings. impaCt of bioLogic therApy on saRs-cov-2 Infection and immuniTY (CLARITY) IBD is a UK-wide, multicentre, prospective observational cohort study investigating the impact of infliximab and vedolizumab and/or concomitant immunomodulators (azathioprine, mercaptopurine and methotrexate) on SARS-CoV-2 acquisition, illness and immunity in patients with IBD.
Study methods have been previously described 10,11 . Consecutive patients were recruited at the time of attendance at infusion units between 22 September 2020 and 23 December 2020 (Supplementary Table 1). Patients aged 5 years and over, with a diagnosis of IBD, treated with infliximab or vedolizumab were eligible for inclusion. Follow-up visits coincided with biologic infusions and occurred eightweekly. Here, we report vaccine-induced antibody responses after the second dose of either the BNT162b2 or ChAdOx1 nCoV-19 vaccines. Participants were eligible for our primary immunogenicity analysis, if they had had an anti-S RBD antibody test between 14 and 70 days after a second-dose vaccine, defined as the second dose of any of the licenced COVID-19 vaccines, 10-14 weeks after the first dose. Anti-S RBD antibody levels were compared with samples from 605 fully vaccinated adult participants from the Virus Watch study, a household community cohort of 10,000 individuals representative of the UK population of England and Wales recruited between 1 June 2020 to 31 August 2021 19 . Peripheral blood mononuclear cells (PBMC) for T cell experiments were collected from patients 4 to 6 weeks after the first and second dose of vaccine at the time of biologic infusions, at selected sites which could facilitate PBMC extraction within 12 h of venepuncture.
Outcome measures. Our primary outcome was anti-S RBD antibodies 2 to 10 weeks after the second dose of the BNT162b2 or ChAdOx1 nCoV-19 vaccines. Secondary outcomes were: (i) the proportion of participants who seroconverted (ii) anti-spike T cell responses in patients following the first and second dose of vaccines (iii) the durability of vaccine responses (iv) risk of breakthrough infections two or more weeks after two doses of vaccine (v) antibody concentrations and seroconversion rates in patients with PCR or serological evidence of past SARS-CoV-2 infection at, or prior, to the postvaccination serum sample Variables. Variables recorded by participants were demographics (age, sex, ethnicity, comorbidities, height and weight, smoking status, and postcode), IBD disease activity (PRO2), SARS-CoV-2 symptoms aligned to the COVID-19 symptoms study (symptoms, previous testing, and hospital admissions for COVID-19) and vaccine uptake (type and date of primary vaccination). Study sites completed data relating to IBD history (age at diagnosis, disease duration, and phenotype according to the Montreal classifications, previous surgeries, and duration of current biologic and immunomodulator therapy) 10 . We linked our data by NHS number or Community Health Index to Public Health England, Scotland, and Wales archive dates and results of all SARS-CoV-2 PCR tests undertaken and vaccines administered. Data were entered electronically into a purpose-designed REDCap database hosted at the Royal Devon and Exeter NHS Foundation Trust 34 . Participants without access to the internet or electronic device completed their questionnaires on paper case record forms that were subsequently entered by local research teams.

Laboratory methods.
To determine antibody responses specific to vaccination we used the Roche Elecsys Anti-SARS-CoV-2 spike (S) immunoassay 35  i. The intra-assay and inter-assay coefficient of variation were 1.3% and 5.6%, respectively ii. Anti-SARS-CoV-2 (S) antibodies were stable in uncentrifuged blood and serum at ambient temperature for up to seven days permitting postal transport iii. No effect was observed on recovery of anti-SARS-CoV-2 (S) antibodies following four freeze/thaw cycles iv. No analytical interference was observed for the detection of anti-SARS-CoV-2 (S) with infliximab or vedolizumab up to 10,000 and 60,000 mg/L, respectively, or with anti-drug antibodies to infliximab or vedolizumab up to 400 and 38 AU/mL, respectively (data not shown). Seroconversion was defined at a threshold of 15 U/mL. ElecSys Anti-SARS-CoV-2 spike (S) RBD concentrations of greater than or equal to 15 U/ml are associated with neutralisation of ≥20% with a positive predictive value of 99.10% (95% CI: 97.74-99.64) 11 .
At the entry to CLARITY IBD and at follow-up visits, all patients were tested for previous SARS-CoV-2 infection using the Roche Elecsys anti-SARS-CoV-2 (N) immunoassay. We have previously reported that anti-N antibody responses following SARS-CoV-2 natural infection are impaired in patients treated with infliximab or vedolizumab 11 . As such, a threshold 0.12 times above the cut-off index was set, using receiver operator characteristic curve and area under the curve analysis of anti-N antibody results from participants two weeks following a PCRconfirmed infection to maximise specificity, beyond which patients were deemed to have had prior SARS-CoV-2 infection (Supplementary Fig. 6). Patients with a PCR test confirming SARS-CoV-2 infection at any time prior to vaccination were deemed to have evidence of past infection irrespective of any antibody test result. Breakthrough infections were defined by a positive SARS-CoV-2 PCR test 2 or more weeks after the second vaccine dose.
Peripheral blood mononuclear cell isolation. Whole blood was collected in lithium heparin tubes and PBMCs were isolated by density-gradient centrifugation using Lymphoprep TM (Stem Cell Technologies) layered onto SepMate TM (Stem Cell Technologies) tubes. PBMC isolation was performed within 12 h of venepuncture. Purified PBMCs were cryopreserved in 10% DMSO/50% FBS and stored in liquid nitrogen pending batch analysis.
Spike-peptide specific T cell responses. IFN-γ T cell ELISpot assays were performed using pre-coated plates (Mabtech 3420-2APT) and using the protocol described previously 28,32 . Two-hundred thousand cells were seeded per well and cells were stimulated with a peptide pool, containing 18 peptides derived from SARS-CoV-2 spike protein 37 at a concentration of 10 μg/ml/peptide; the peptide pool utilises a mapped epitope pool (MEP) or 12-20mer peptides, mapped as eliciting high-prevalence CD4 responses covering diverse HLA-II haplotypes 28,32 . Use of this spike MEP in otherwise healthy SARS-CoV-2 seropositive individuals elicits a T cell response in 83% of individuals at 16-18 weeks after natural SARS-CoV-2 infection and 91% of healthy individuals 2-3 weeks after two-dose vaccination with seronegative individuals showing a level of response indistinguishable from pre-pandemic controls 28,32 . Plates were cultured for 18-20 h before development and data were collected using an AID classic ELISpot plate reader (Autoimmun Diagnostika GMBH). Results are expressed as differences in (delta) spot forming cells (SFC) per 10 6 PBMC between peptide stimulation and a mediaonly control. A response below 2 standard deviations of the media-only control wells was deemed to be a null response. Data were excluded if the response to the positive control anti-CD3 stimulation was <200 SFC per 10 6 PBMCs.  Sample size. The sample size for CLARITY IBD was based on the number of participants required to demonstrate a difference in the impact of infliximab and vedolizumab on seroprevalence and seroconversion following SARS-CoV-2 infection, with an estimated background seroprevalence of 0.05. We calculated that a sample of 6970 patients would provide 80% power to detect differences in the seroprevalence of SARS-CoV-2 antibodies in infliximab-treated compared with vedolizumab-treated patients, whilst controlling for immunomodulator status at the 0.05 significance level.
Statistical analyses. Analyses were undertaken using R 4.1.2 (R Foundation for Statistical Computing, Vienna, Austria). All tests were two-tailed and p values were reported without correction for multiple testing. P values <0.05 were considered significant. We included patients with missing clinical data in analyses for which they had data and have specified the denominator for each variable. Anti-S RBD antibody concentrations are reported as geometric means and standard deviations. Other continuous data are reported as a median and interquartile range, and discrete data as numbers and percentages, unless otherwise stated.
Univariable analyses, using Spearman's rank correlation coefficients, and t-tests of log-transformed anti-S RBD antibody concentration were used to identify demographic, disease, vaccine and treatment-related factors associated with the concentration of anti-S RBD antibodies across the cohort. Crude sensitivity analyses excluding patients treated without a concomitant immunomodulator were undertaken to control for the effect of immunomodulator use on anti-S RBD antibody concentrations. Propensity matching was used to account for the other significant differences in baseline variables between infliximab-treated and vedolizumab-treated patients using the MatchIt package 38 . A priori, patients were matched exactly on diagnosis, immunomodulator use, and then using optimal matching, on age, the number of comorbidities, ethnicity, and presence of active disease. Multivariable linear regression models were used to identify factors independently associated with log anti-S RBD concentration. A priori, we included age, ethnicity, biologic medication and immunomodulator use. Results are presented after exponentiation so that the coefficients of the model correspond to the fold change (FC) associated with each binary covariate. For age, a cut-off was chosen based on a graphical inspection of the relationship between age and anti-S RBD antibody concentrations.
Mann-Whitney U-test was used to compare the magnitude of T cell response (SFC/10 6 PBMCs) stratified by treatment and vaccine received, and Spearman's rank correlation coefficient was calculated to determine the correlation between antibody and T cell responses.
Anti-S RBD antibody half-lives were estimated using an exponential model of decay. Linear mixed models were fit using the lme4 and lmerTest package, with biologic treatment and vaccine type as fixed effects and each subject as a random effect. Each of these effects were estimated independently for gradient and intercept. 95% confidence intervals of fixed effects were calculated using likelihood ratios. P values for comparison of half-lives were estimated from the full linear mixed-effects model that incorporated vaccine, biologic drug and prior SARS-CoV-2 infection status.
We visualised the durability of antibody responses by calculating 15-day rolling geometric mean anti-S RBD antibody concentrations. For this analysis we included participants who had an antibody test carried out between 1 and 70 days after the second vaccine dose. Cox proportional hazard regression models were used to identify the demographic, disease and treatment-related factors associated with the time to fall in anti-S RBD antibody concentration below the seroconversion threshold.
Kaplan-Meier curves and Cox proportional hazard regression model was used to identify treatment-related factors associated with time to breakthrough infection. A linear regression model of log-transformed geometric mean anti-S RBD antibody concentration was used to determine the risk of breakthrough infections.
Where appropriate the same analyses were used to compare antibody responses in participants with PCR evidence of SARS-CoV-2 infection at any time prior to vaccination.
Ethical consideration. Patients were included after providing informed, written consent and compensation for participation was not provided. The sponsor was the Royal Devon and Exeter NHS Foundation Trust. The Surrey Borders Research Ethics committee approved the study (REC reference: REC 20/HRA/3114) in September 2020. The protocol is available online at https://www.clarityibd.org. The study was registered with the ISRCTN registry (https://doi.org/10.1186/ ISRCTN45176516).
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The study protocol including the statistical analysis plan is available at https:// www.clarityibd.org/. Individual participant de-identified data that underlie the results reported in this article will be available immediately after publication for a period of 5 years. Due to the sensitive nature of the data, this will be made available to investigators whose proposed use of the data has been approved by an independent review committee.
Analyses will be restricted to the aims in the approved proposal. Proposals should be directed to tariq.ahmad1@nhs.net. To gain access data requestors will need to sign a data access agreement. Data from the Virus Watch study is currently being archived on the Office of National Statistics Secure Research Service and will be available shortly. Source data are provided with this paper in the Source Data file. Source data are provided with this paper.

Code availability
Statistical analyses were undertaken in R 4.1.2 (R Foundation for Statistical Computing, Vienna, Austria. Code has been made available at: https://github.com/exeteribd/ clarityibd-public.