The study design and protocol for the COVID-19 PROTECT study group were assessed by the National Healthcare Group (NHG) Domain Specific Review Board (DSRB) and approved under study number 2012/00917. Written informed consent was obtained from all study participants in accordance with the Declaration of Helsinki for Human Research. A cohort of 95 LTRs was recruited for the study. The interval between the first and second dose of BNT162b2 was 21 days (IQR: 21-24 days). On day 180, 61 out of the 95 LTRs had received a third dose of BNT162b2. The interval between the third dose of BNT162b2 and day 180 post first dose was 76 days (IQR: 54.5-97.75 days). The remaining 34 LTRs who did not receive the third dose and were COVID-19 positive before day 180 sampling were excluded. Blood collection was performed on days 0 (i.e., before the first BNT162b2 dose), 90 post first dose, and 180 post first dose. Blood samples from 268 age-matched healthy individuals who received two mRNA vaccines (BNT162b2) on day 90 post first dose were used as a control (HC). None of the individuals had known or reported SARS-CoV-2 infection and change in the IS regimens due to episodes of rejection or side effects of the medications.

The Elecsys^® Anti-SARS-CoV-2 S (Roche S) immunoassays were used to measure antibodies against the receptor-binding domain (RBD) of the spike protein. The Roche Cobas e411 Analyzer (Roche, Basel, Switzerland) was used for the assay according to the manufacturer’s instructions. The Roche S assay measured the electro-chemiluminescent signal representing antibody levels in titrated samples. Samples with antibody levels of ≥ 0.8 U/mL were considered positive.

The SFB assay was performed according to previously described methods (16, 17). In brief, cells expressing the spike protein of the ancestral Wuhan strain, Omicron BA.1 and XBB were seeded at 1.5 x 10⁵ cells/well in 96-well V-bottom plates (Thermo Fisher Scientific, Waltham, USA). The cells were incubated with human serum (diluted 1:100 in 10% FBS; HyClone, Chicago, USA), followed by a second incubation with a double stain comprising Alexa Fluor 647-conjugated anti-human IgG (diluted 1:500; Thermo Fisher Scientific) and propidium iodide (PI; diluted 1:2500; Sigma-Aldrich, Burlington, USA). Cells were acquired using an LSR4 laser (BD Biosciences, New Jersey, USA) and analyzed using FlowJo (Tree Star, BD Biosciences). The percentage of GFP-positive spike protein-expressing cells bound by the antibody, indicated by Alexa Fluor 647- and FITC-positive events, was used as an indicator of binding. The assay was performed as two independent experiments, each with technical duplicates. The amount of spike protein expressed on the cell surface was verified by ACE-2-HuFc binding. A subset of age-matched samples was randomly selected and examined for binding antibodies against Omicron BA.1 and XBB (n = 10 per LTRs with double and triple IS regimens and HC).

SARS-CoV-2 RBD-specific memory B cell (MBC) numbers were quantified using the ELISpot Path: Human IgG (SARS-CoV-2, RBD) ALP kit (Mabtech, Cincinnati, USA) following the previously described protocol (18). PBMCs were suspended in RPMI + 10% FBS + 1 μg mL^-1 R848 + 10 ng mL^-1 IL-2 and incubated at 37°C for 5 days to allow for differentiation into antibody-secreting cells. To determine RBD-specific MBC numbers, 100,000 or 400,000 live cells were plated for ELISpot. Total IgG-secreting cells were detected by plating 1,500 or 3,000 live cells to normalize the results. Plates were then read on an IRIS ELISpot reader (Mabtech), and spots were quantified based on the average of duplicate wells. Due to limited cell availability, a subset of age-matched samples was randomly selected and examined (n = 42 for LTRs; n = 52 for HC).

SARS-CoV-2-specific T cell subsets were characterized using a previously described method with modifications (19, 20). PBMCs were rested overnight in RPMI-1640 + 5% human serum at 37 ˚C then stimulated with PMA (100ng/mL) (Sigma Aldrich) and ionomycin (1µg mL^-1) (Sigma Aldrich) as a positive control or with pooled SARS-CoV-2 PepTivator^® S and S1 peptides (0.6nmol mL^-1 each) (Miltenyi Biotec), or left unstimulated (baseline) for 6 h. Brefeldin A and Monesin (Thermo Fisher Scientific) were added at 2 h post stimulation. Cells were stained with surface markers for 30 min ( Table S1 , #1 to 21), fixed and permeabilized with Foxp3/Transcription Factor Staining Buffer Set (Thermo Fisher Scientific) for 30 min, and then stained for intracellular cytokines for another 30 min ( Table S2 , #22 to 29). Cells were acquired with the CytekTM Aurora (SpectroFlo^®) and analyzed using FlowJo. Spike-specific intracellular granzyme B expression was determined after baseline subtraction. A subset of age-matched samples was randomly selected and examined due to limited cell availability (n = 41 for LTRs; n = 40 for HC).

The pseudotyped lentivirus neutralization assay was performed according to a previously described protocol with slight modifications (21). Briefly, CHO-ACE2, a stable cell line expressing human ACE2, was acquired from Associate Professor Dr Yee-Joo Tan (Department of Microbiology, National University of Singapore & Institute of Molecular and Cell Biology, A*STAR, Singapore) (22) and utilized for the assay. The CHO-ACE2 cells were seeded at 1.8 x 10⁴ cells per well in a 96‐well black microplate (Corning, New York, USA) with DMEM without Geneticin and were allowed to settle overnight. Heat‐inactivated plasma samples were serially diluted (1:5 to 1:5120 dilutions) and incubated with an equal volume of pseudovirus-expressing spike proteins of the respective SARS-CoV-2 strain (5 ng of p24 per well) at 37°C for 1 h. The mixture was then added in duplicate to the pre‐seeded CHO‐ACE2 cells. The wells were topped up with DMEM after 1 h of incubation. After 48 h, cells were washed with phosphate-buffered saline (PBS) and lysed with 1X Passive Lysis Buffer (Promega) with gentle shaking at 125 rpm at 37°C for 30 min. Luciferase activity was subsequently quantified using the Luciferase Assay System (Promega) on a GloMax Luminometer (Promega). IC50 values were calculated as the reciprocal of the dilution at which a 50% reduction in luciferase activity was observed. A subset of age-matched samples was randomly selected and examined for binding antibodies against the spike protein of the ancestral Wuhan strain, as well as Omicron BA.1 and XBB (n = 10 per LTRs with double and triple IS regimens and HC).

Statistical analysis was performed using GraphPad Prism 7. To compare different time points, unpaired comparisons were performed using the Mann-Whitney U-test. For the comparison between LTRs with different regimens and HC, the Kruskal-Wallis test was used, followed by post hoc tests. Dunn’s tests were used to correct for multiple comparisons. To compare between matched samples, the Wilcoxon matched-pairs signed-rank test was used. All tests were two-tailed, and p < 0.05 was considered statistically significant.

A cohort of 95 LTRs receiving various IS treatments ( Table S1 ) were vaccinated with the Pfizer/BioNTech vaccine, BNT162b2, and their immune response was monitored. As LTRs are known to mount reduced antibody response (14, 23), the third dose of BNT162b2 was recommended for them. On day 180, 61 out of the 95 LTRs had received a third dose of BNT162b2, with a median interval of 76 days since the third dose ( Table S1 ).

After two doses of BNT162b2 (day 90), antibodies to full-length spike and RBD were induced and significantly increased in all LTRs, although the levels were significantly lower than those in HC on day 90 ( Figures 1A, B ). Following the third dose of BNT162b2, the level of antibodies against the full spike protein increased in LTRs but remained lower than in HC who received two doses ( Figure 1A ). In contrast, the level of antibodies against the RBD increased and was significantly higher than in HC who received two doses ( Figure 1B ).

Successful induction of humoral immune responses requires a well-coordinated response of B and T cells (24), with an effective CD8 cytotoxic T cell response being critical for eliminating the virus (25). To determine whether BNT162b2 induces spike-specific B and T cell recall responses, we next examined a subpopulation of vaccinated LTRs (n=42) due to limited cell availability. After two doses of BNT162b2, LTRs had significantly lower RBD-specific MBC response compared to HC vaccinated with two doses. However, after receiving the third dose of BNT162b2, the RBD-specific MBC response in LTRs increased and was similar to the response in HC vaccinated with two doses ( Figure 1C ).

To assess CD8+ T cell response, we measured the percentage of CD8+ T cells secreting granzyme B after SARS-CoV-2 peptide stimulation. After two doses of BNT162b2, the granzyme B-producing CD8 T cell response in LTRs remained significantly lower than in HC vaccinated on day 90. However, following the third dose of BNT162b2, a significant increase in the level of spike-specific granzyme B-producing CD8 T cells was observed in LTRs, becoming similar to the response of HC vaccinated with two doses ( Figure 1D ). Overall, LTRs required three doses of the vaccine to show robust antibody, MBC, and CD8+ T cell responses comparable to those seen in HC.

Calcineurin inhibitors are the most commonly prescribed IS drugs for maintenance immunosuppression in transplant recipients (26). However, there have been concerns regarding nephrotoxicity in transplant recipients when given at high doses (27). As a result, mycophenolate mofetil (MMF) and other IS drugs have been prescribed to replace or minimize the use of high calcineurin inhibitor doses in LTRs (28–30). Previous studies have demonstrated that an IS regimen with three different drugs or the use of MMF affects vaccine immunogenicity in LTRs (9, 11). Therefore, we stratified LTRs based on the number of received IS regimens and evaluated any differences in humoral and cellular responses after vaccination.

We compared the different IS regimens after two vaccine doses and found that LTRs receiving a single IS regimen had significantly higher spike- and RBD-specific IgG antibodies than the stratified LTRs receiving multiple IS regimens. Spike- and RBD-specific IgG antibody levels of single IS LTRs did not differ from vaccinated HC. In contrast, LTRs receiving double and triple IS regimens had significantly lower spike- and RBD-specific IgG antibodies than vaccinated HC ( Figures 2A, B ). Moreover, LTRs receiving the triple IS regimen had significantly lower neutralizing antibodies against ancestral Wuhan spike compared to HC vaccinated with two doses ( Figure 2C ). Regarding MBC response, single and double IS regimens achieved an MBC response similar to vaccinated HC, while the triple IS regimen had a significantly lower MBC response ( Figure 2D ). For CD8+ T cell response, the percentage of granzyme B-producing CD8+ T cells was similar between the single IS regimen and HC. Only the double IS regimen had significantly lower levels of granzyme B-producing CD8 T cells compared to HC vaccinated with two doses ( Figure 2E ). Thus, LTRs receiving a single IS regimen did not show significantly different immune responses compared to HC, whereas LTRs receiving double and triple IS regimens had lower antibody, MBC, and CD8+ T cell responses.

Next, we aimed to investigate whether a third dose of BNT162b2 could rescue the reduced immune responses observed in LTRs with double and triple IS regimens. After three doses of BNT162b2, LTRs with the double IS regimen had a similar level of antibodies against the full spike protein to HC with two vaccine doses, whereas LTRs with the triple IS regimen continued to have a significantly lower level of antibodies to the full spike protein compared to vaccinated HC ( Figure 3A ). Against the RBD protein, LTRs with double and triple IS regimens showed similar antibody levels to HC vaccinated with two doses ( Figure 3B ). Furthermore, the percentage of LTRs receiving multiple IS regimens who were antibody non-responders decreased 2-fold (from 22.86% to 0%) and 3-fold (from 53.3% to 10%) after three doses of BNT162b2 ( Figure 3C ). In addition, for LTRs receiving both double and triple IS regimens, neutralizing antibody responses were improved and were not significantly different compared to HC vaccinated with two doses of BNT162b2 ( Figure 3D ).

Similarly, after three doses of BNT162b2, both double and triple IS regimens showed similar MBC response to HC vaccinated with two doses, with no significant difference observed ( Figure 3E ). Additionally, the CD8+ T cell response in LTRs receiving double or triple IS regimens was boosted, with no significant difference in the percentage of granzyme B-producing CD8 T cells compared to HC vaccinated with two doses ( Figure 3F ). LTRs receiving a single IS regimen had significantly higher RBD-specific MBC response on day 180 than LTRs with multiple IS regimens or HC ( Figure 3E ). Although differences were observed in B cell response, all other LTR groups had comparable granzyme B-producing CD8 T cell responses after the third dose of BNT162b2 irrespective of the IS regimen ( Figure 3F ). Thus, a third dose of BNT162b2 rescued the reduced immune responses observed in LTRs with multiple IS regimens, leading to enhanced antibody, MBC, and CD8+ T cell responses that were similar to those observed in HC vaccinated with two doses.

Reduced antibody response has been observed against the emergent Omicron variant and its sublineages compared to the ancestral Wuhan strain in immunocompetent individuals (31, 32). Therefore, we aimed to investigate the cross-variant antibody breadth in LTRs receiving double and triple IS regimens after the third dose of the BNT162b2 vaccine against Omicron BA.1 and XBB variants.

Binding antibody responses against Omicron sublineages increased after the third dose of BNT162b2 for LTRs receiving both double and triple IS regimens ( Figures 4A, B ). Interestingly, LTRs with the double IS regimen showed higher antibody response against Omicron BA.1 than HC vaccinated with two doses ( Figure 4C ). Similarly, neutralizing antibodies were boosted against Omicron BA.1 ( Figure 4D ), although neutralization activity against XBB remained very low ( Figure 4E ). Notably, LTRs with double but not triple IS regimens showed higher antibody response against Omicron BA.1 than HC vaccinated with two doses ( Figure 4F ). Thus, the three-dose regimen in LTRs provided a cross-variant antibody breadth similar to that of HC vaccinated with two doses of BNT162b2.