Blood and serum samples were collected from 48 HSCT patients and controls prospectively enrolled between March 2021 and February 2022 to examine the immune response following two doses of (mRNA) vaccines Moderna/mRNA-1273 or Pfizer/BioNTech’s BNT162b2. All HSCT patients were negative for IgG anti-RBD Nabs (BAU/mL) before vaccination. HSCT patients had all received allotransplants and 26 were recruited from the graft v host disease (GvHD) clinic where patients that develop GvHD receive prolonged immunosuppression. This group was expected to have weaker vaccine responses (see Supplementary Table S1 ).

Ex vivo vaccine-dependent T cell reactivity of HSCT patients was compared to the reactivity of T cells from 18 healthy controls (14 females and 4 males) who had no history of COVID-19 infection and were all negative for IgG anti-nucleocapsid. The median time of blood collection after the last dose was 6 weeks. All procedures followed here were carried on in accordance with institutional and national ethical standards of the responsible committee on human experimentation and with the 2013 Helsinki Declaration. Informed consent was obtained from all study participants and approved by the Health Region South-East Regional Ethics committee.

Spike-C (PepTivator® SARS-CoV-2 Prot_S Complete; Miltenyi, # 130-127-953), and Spike-I (PepTivator® SARS-CoV-2 Prot_S; Militenyi, # 130-126-700) pools are collections of lyophilized 15-mers peptides with 11 aa overlaps spanning either the immunodominant regions (Spike-I pool sequence domains: aa 304-338, 421-475, 492-519, 683-707, 741-770, 785-802, and 885-1273), or the entire length (Spike-C pool sequence: aa 5-1273) of the SARS-CoV-2 Spike glycoprotein (Protein QHD43416.1, GenBank MN908947.3). NOI pools are collections of 9 to 10-mers peptides identified by the NEC Immune Profiler algorithm (25). The pools include the following peptides: AHFPREGVF, ASFSTFKCY, DKVFRSSVL, ESIVRFPNI, FAMQMAYRF, FKNLREFVF, FVFKNIDGY, FVSNGTHWFV, GTHWFVTQR, GAAAYYVGY, IPFAMQMAY, IPTNFTISV, KTSVDCTMY, KVGGNYNYLY, LPFNDGVYF, LPIGINITRF, LPPAYTNSF, NSFTRGVYY, RGWIFGTTL, RLITGRLQSL, RAAEIRASA, SIIAYTMSL, SPRRARSVA, TPINLVRDL, TRFASVYAW, TRFQTLLAL, WTAGAAAYY, YLQPRTFLL, YQPYRVVVL, YTNSFTRGVY. Peptides were synthesized by GenScript (Piscataway, NJ, USA) at a purity ≥85%, and pooled at a final concentration of 1.5mg/mL each.

PBMC were isolated from whole blood using either CPT tubes (BD vacutainer, # 362782) or Lymphoprep™ (Serumwerk Berburg, # 1858). Blood in CPT tubes was first centrifuged at 1600g for 25 minutes at room temperature and plasma collected and stored at -20°C. PBMC were then transferred into a 50mL tube, washed in cold PBS (Gibco, # 10010-015), and counted. Lymphoprep™ isolation was performed per manufactured protocol. Briefly, blood was pipetted into a 50mL tube, mixed 1:1 at room temperature with PBS, and layered onto 10 to 15mL of Lymphoprep™ in a 50mL falcon tube. Tubes were centrifuged at 800g for 25 minutes and PBMC collected, washed 3 times with PBS (400g, 7 min, 4°C) and resuspended in FBS (Gibco, # 10270-106) complemented with 10% DMSO. After overnight pre-chilling at -80°C in Mr. Frosty (Nalgene™, # 5100-0001) cells were transferred in liquid nitrogen for long-term storage.

Whole blood was collected in 5mL Vacuette® tubes with serum separator clot activator (Vacuette, # 456073R). Serum was obtained by centrifugation at 2000g and stored at -20°C. Semiquantitative measurement of antibodies to full-length spike protein (Spike-FL) and the receptor-binding domain (RBD) from SARS-CoV-2 was performed using a multiplexed bead-based assay (26). Polymer beads with fluorescent barcodes were coupled successively to neutravidin (Thermo Fisher) and biotinylated viral antigens to generate bead-based protein arrays. Sera were diluted 1: 100 in assay buffer (PBS 1% Tween 20, 10μg/mL D-biotin, 10μg/mL Neutravidin, 0.1% Sodium Azide). Diluted sera were incubated with bead-based arrays in 384 well plates for 30 minutes at 22°C under constant agitation, washed three times in PBS 1% Tween 20 (PBT) and labelled with R-Phycoerythrin (R-PE)-conjugated goat-anti-human IgG (Jackson Immunoresearch). For measurement of neutralizing antibodies, the beads were pelleted after incubation with serum and labelled successively with digoxigenin-conjugated human ACE2 and mouse monoclonal anti-dixogigenin (Jackson Immunoresearch), which was conjugated in-house to R-PE. The beads were analyzed with an Attune NxT flow cytometer (Thermo Fisher), and raw data (fcs.3.1) were analyzed in WinList 3D (Verity Softwarehouse). The median R-PE fluorescence intensity (MFI) of each bead subset was exported to Excel. The MFI of beads coupled with viral antigens was divided by that measured on beads coupled with neutravidin only (relative MFI, rMFI). A total of 979 prepandemic sera and 810 sera from COVID-19 convalescents were analyzed to establish cut-offs for seropositivity. A double cut-off of rMFI >5 for anti-RBD and anti-Spike FL yielded a specificity of 99.7% and a sensitivity of 95% (27). Serum from an individual who had received three doses of the Pfizer/BioNTech anti-COVID-19 vaccine was used as standard to convert signals to binding antibody units per millilitre (BAU/mL).

PBMC were first washed in FACS Wash Buffer [PBS1X w/o Ca++ and Mg++ (Gibco # 10010023) supplemented with 1% bovine serum albumin (VWR, #K719)] and stained for 10 minutes in the dark in 150μL of cold PBS containing Fixable Near IR Live/Dead viability stain (Molecular Probes, # L34976) for dead cell exclusion. Cells were then washed once in cold PBS1x and incubated for 30 minutes on ice in 50µL of FACS Wash Buffer containing the following fluorochrome-conjugated antibodies: BV605 anti-human CD3 (Clone SK7; BD Biosciences, # 563219), PerCP-Cy5.5 anti-human CD4 (Clone OKT4; Biolegend, # 317428), and Alexa Fluor 488 anti-human CD8 (Clone OKT8; Invitrogen, # 53-0086-42). After incubation and washing, cells were permeabilized in 150µL BD Cytofix/Cytoperm solution for 30 minutes at room temperature and then washed twice in 200µL of 1x BD perm/wash solution (BD Biosciences Fixation and permeabilization kit; # 554714). Intracellular staining was performed in 50µL of 1x BD perm/wash solution containing the following fluorochrome-conjugated antibodies: APC anti-human CD137 (Clone 4B4-1; BD Biosciences, # 550890), BV711 anti-human CD40L (Clone 24-31; Biolegend, # 310837), PE anti-human IFN-γ (Clone 4S.B3; Biolegend, # 502509), and BV421 anti-human TNFA (Clone MAb11; BD Biosciences, # 562783) according to manufactures instructions. Cells were washed one more time in PBS, resuspended, and acquired on an Attune NxT Flow Cytometer (Thermo Fisher).

Quantification of T cell activation was performed by flow cytometry as described (28, 29). Briefly, PBMC were thawed, washed twice, and resuspended in RPMI 1640 medium with GlutaMAX™ supplement (Thermo Fisher Scientific, # 61870-010), 1mmol/L Sodium Pyruvate (Gibco # 11360-039), 1mmol/L MEM NEAA (Gibco # 11140-035), 50nmol/L 1-thioglycerol (Sigma-Aldrich, # M1753), 12μg/mL Gentamycin (VWR, # E737), and 10% heat-inactivated Foetal Bovine Serum (Gibco, # 10270-106). Cells were plated in a 96-well round bottom cell culture plate (1M cells in 200μL per well), stimulated for 3 hours with Spike-I or Spike-C pool solutions (8μl), or NOI pools (1.5μg/mL final concentration per peptide). Cells were then further incubated for 18h after addition of Brefeldin A/Monensin cocktail (GolgiStop 500X, Invitrogen # 00-4980-93). At the end of the incubation, the plate was placed on ice, and cell pellets washed once in cold PBS before flow cytometry processing.

To evaluate T cell reactivity to antigenic challenge, a Reactivity Score (RS) metric was generated from the analysis of cell populations defined by the combined expression of 4 AIMs. These markers and 3 gated areas of interest (Single +, Double +, and “All”) were used to define a total of 16 partially overlapping cell populations ( Supplementary Figure S1A , Supplementary Table S2 ). We selected T cell populations whose response (frequency after stimulus minus background frequency) was significantly higher in healthy controls than HSCT patients and normalized each value by the average frequency of all measurements for that population. When the frequency after stimulation was smaller than the background (frequency after stimulus minus background frequency < 0), the patient was assigned a RS value of ‘zero’. The RS for each patient was then computed by averaging the normalized frequency of the selected populations for each T cell subset (selected T cell populations are shown in Supplementary Figure S2 ). Similarly, the score for differential reactivity (DR) was computed by normalizing the difference in T cell response to Spike-C and Spike-I pool stimulation (Spike-C minus Spike-I). The T cell populations selected for DR calculation are shown in Supplementary Table S2 . Raw frequency data for each population are reported in Supplementary Table S3 .

Statistical analyses were performed using GraphPad Prism V.8 (GraphPad software). Two-tailed Fisher’s Exact test, Binomial test, Wilcoxon matched pairs signed rank test, and Mann-Whitney test were used when appropriate. The Spearman’s rank correlation coefficient was computed to assess the correlation between variables.

Peripheral blood mononuclear cells (PBMC) and serum samples were collected between March 2021 and February 2022 from 48 HSCT recipients 4 weeks after receiving two doses of Moderna/mRNA-1273 or Pfizer/BioNTech BNT162b mRNA vaccines (see Supplementary Table S1 for an overview of the patients). The median time from transplantation was 435 d [IQR, 273-975]. Of the 48 HSCT patients, 26 received continued immunosuppression 6 months after transplantation due to GvHD (a high rate, as many patients were recruited from the GvHD clinic). As of January 2023, only one patient developed severe COVID-19 requiring hospitalization with mechanical ventilation after Delta VOC infection.

Following vaccination, the concentration of serum IgG anti-RBD in HSCT patients was significantly lower than in healthy controls (P < 0.0001; Figure 1A ) with a median IgG anti-RBD level of 898 BAU/mL [IQR 10-5842] in the HSTC group, and 6677 BAU/mL [IQR 4358-8395] in controls. As shown in Figure 1A , only 37.5% (18/48) of patients had protective levels of IgG anti-RBD (>2000 BAU/mL), 21% (10/48) had a clear response (200-2000 BAU/mL), 10% (5/48) had low response (20-200 BAU/mL), while the remainder 31% (15/48) did not seroconvert (0-20 BAU/mL). We found a positive correlation between IgG anti-RBD levels and the time after transplantation (Spearman r = 0.32, P = 0.027; Figure 1B ), but no correlation between the magnitude of response and gender or age ( Figures 1C, D ). This cohort included patients with high risk of severe COVID-19 who were receiving immunosuppression for GvHD. Nevertheless, we found no significant difference in response between recipients of cells from HLA-identical siblings and those from unrelated donors ( Figure 1E ), and no impact of immunosuppressive therapy on vaccine response (Fisher’s Exact, P = 0.236; Figure 1F and data not shown). Notwithstanding the magnitude of response after the second dose, 8 of 12 double-vaccinated patients (67%) for whom longitudinal serology data were available benefited from a 3^rd and a 4^th dose, 1 patient showed suboptimal response after the 3^rd dose, and 3 non-responders (25%) failed to seroconvert even after the 3^rd or 4^th dose ( Figure 1G ). In this sub-cohort, the median time for serum IgG assessment after vaccination was 4.5, 8, and 5 weeks for the 2^nd, 3^rd, and 4^th dose, respectively. Of note, 3 of these 8 patients showed positive serological results for the IgG anti-nucleocapsid (not shown), indicating that the higher IgG anti-RBD levels measured were the likely result of hybrid immunity. These non-responders had developed severe GvHD or relapse of malignancy.

We first examined spike peptide-specific T cell responses towards a pool of immunodominant SARS-CoV-2 peptides (Spike-I pool, see Methods). To measure the response, we analyzed the activation of CD4⁺ Th cells and CD8⁺ CTL using activation-induced markers (AIMs) gated as illustrated in Supplementary Figure S1A and Methods. We found that the CD4⁺/CD8⁺ T cell ratio was significantly lower than in healthy donors (0.265 vs. 0.894 median values; P < 0.0001; Supplementary Figure S1B ), a common observation in patients that have undergone bone marrow transplantation. Of 34 HSCT patients tested for T cell reactivity, 56% (19/34) had increased Spike-I specific CD4⁺ Th cells, as measured by CD137 and CD40L markers ( Figure 2A ), a response significantly lower than in controls where 18/18 responded (P = 0.019; Figure 2E ). Similarly, we found a lower frequency of Spike-I specific TNF⁺ IFN-γ⁺ CD4⁺ Th cells ( Figure 2B ) with responses detected in only 8.3% of patients analyzed with this marker combination (P = 0.0004; Figure 2F ) as well as a lower response of Spike-I specific CD8⁺ T cells expressing CD137 and IFN-γ ( Figure 2C ), or CD137 and TNF ( Figure 2D ), which were observed in 43% (18/42) (P = 0.015; Figure 2G ), and 28% (8/29) (P = 0.028; Figure 2H ) of patients, respectively. Because we found some individual variation in the response patterns of the CD137, CD40L, TNF, and IFN-γ markers in both CD4⁺ and CD8⁺ T cell subsets, we used the reactivity score (RS) metric, a normalized value that incorporates the contribution of all relevant combinations of these 4 AIMs to provide a more unbiased quantification of reactivity (see methods and Supplementary Figure S1A ) and confirmed that overall Spike-I specific response was significantly reduced in both the CD4⁺ and CD8⁺ T cell compartment of HSCT patients (see patients’ ranking by RS in Supplementary Figures S2A, B ). Accordingly, comparison of RS values confirmed lower reactivity in HSCT patients for both the CD4⁺ (median 0.842 vs 0.315, P = 0.0031; Figure 2I ) and CD8⁺ T cell subset (median 0.62 vs 0.32 value; P = 0.0046; Figure 2J ). Nonetheless, CD4⁺ and CD8⁺ T cell response to the epitopes from the immunodominant regions of the S protein remained undetectable in 26% and 10% of HSCT patients, respectively.

To ensure that we did not overlook any T cell specificities, we also tested overlapping peptides that covered the complete spike protein (Spike-C pool). Surprisingly, the response of HSCT recipients to wider antigenic stimulation was much improved compared to the immunodominant spike peptide pool. In fact, HSCT recipients and controls were not significantly different in terms of changes in frequency of CD137⁺ TNF⁺ and CD137⁺ IFN-γ⁺ CD4⁺ T cell (P = 0.7; Figures 3A, B ) or CD137⁺ and IFN-γ⁺ CD8⁺ T cell populations (P > 0.4; Figures 3C, D ). We observed better response to Spike-C stimulation for CD137⁺ TNF⁺ (P = 0.0003, Wilcoxon ranked test; Figure 3E ) and CD137⁺ IFN-γ⁺ (P = 0.0008; Figure 3F ) CD4⁺ cells, and an increase in the fraction of patients with detectable response (90% and 85% for the two cell populations, respectively; P < 0.0001 and P = 0.0002, Binomial test; Figures 3E, F ). CD8⁺ T cells showed similar trends, with better response to Spike-C for cells expressing CD137⁺ alone (P = 0.001; Figure 3G ) and in combination with IFN-γ⁺ (P = 0.0005; Figure 3H ), and a response rate of 95% and 91%, respectively (P < 0.0001, Binomial test; Figures 3G, H ). Altogether, these results suggest that after vaccination HSCT recipients developed cell-mediated immunity with different T cell specificities.

To compare differences in stimulus-specific T cell response (T cell specificity) between cohorts we defined a score that reflects differential reactivity (DR) to Spike-I and Spike-C stimulation (see Methods for details). A greater fraction of HSCT recipients showed DR values higher than controls for both CD4⁺ ( Figure 4A ) and CD8⁺ ( Figure 4B ) subsets indicating that the response to Spike-C was generally stronger in patients than controls. The HSCT cohort had higher DR for CD4⁺ T cells (median DR = 0.55 vs. 0.090; P = 0.018; Figure 4C ), but not for CD8⁺ T cells (median DR = 0.78 vs. 0.14; P = 0.065; Figure 4D ). Likewise, the ratio of the RS of Spike-C and Spike-I (Spike-C RS/Spike-I RS) was significantly higher in the HSCT cohort than controls for CD4⁺ T cells (median ratio of 1.1 vs. 0.68; P = 0.0011; Figure 4E , left), but not CD8⁺ T cells (1.3 vs 0.91, P = 0.055; Figure 4E , right), indicating a difference in specificities between the two T cell subsets. This distinction could be explained in part by the fact that the pools are composed of 15-mers peptides which preferentially activate CD4⁺ T cells. For a closer analysis of CD8⁺ T cells, we tested response to the NEC OncoImmunity (NOI) pool, a collection of 9-10-mers S protein epitopes optimized for the detection of spike-specific CD8⁺ T cells - see Methods and (25). HSCT patients had reduced CD4⁺ T cell reactivity after stimulation with the NOI pool (0.34 vs. 0.12; P = 0.012; Figure 4F ), but the response of the CD8⁺ T cell subset was retained (0.40 vs. 0.27; P = 0.4; Figure 4G ). Altogether, these results suggest that the skewed peptide specificity profile within the CD4⁺ T cell compartment is unique to HSCT recipients.

Next, we investigated the relationship between humoral response and T cell immunity in the context of both Spike-I and Spike-C stimulation. Seroconverted and non-responders had similar reactivity to Spike-I for both CD8⁺ (median RS of 0.40 vs. 0.21; P = 0.26; Figure 5A ) and the CD4⁺ T cells (median RS of 0.40 vs. 0.094; P = 0.12; Figure 5B ). Moreover, time from transplantation was predictive of T cell reactivity for CD8⁺ (Spearman’s r = 0.33, P = 0.034; Figure 5C ) and CD4⁺ T cells (Spearman’s r = 0.34, P = 0.047; Figure 5D ), even though there was no association between the responses of the two T cell subsets (Spearman’s r = 0.28, P = 0.11, ns; Figure 5E ). Moreover, IgG anti-RBD level correlated with the reactivity of CD4⁺ (Spearman’s r = 0.40; P = 0.020; Figure 5F ), but not CD8⁺ T cells ( Figure 5G ). Likewise, T cell reactivity in seroconverted and non-responders following Spike-C stimulation were not significantly different in both the CD8⁺ (median RS of 0.65 vs. 0.18; P = 0.095; Figure 5H ), or the CD4⁺ T cell subset (median RS of 0.43 vs. 0.16; P = 0.067; Figure 5I ). However, Spike-C stimulation revealed a strong correlation between the time after transplantation and CD4⁺ T cell reactivity (Spearman’s r = 0.66; P = 0.0014; Figure 5J ), which was not observed for the CD8⁺ T cell compartment ( Figure 5K ). We found no correlation between CD4⁺ and CD8⁺ T cell reactivity to Spike-C ( Figure 5L ), while IgG anti-RBD level correlated with the reactivity of CD4⁺ T cells (Spearman’s r = 0.55; P = 0.013; Figure 5M ), but not CD8⁺ T cells ( Figure 5N ). Altogether, these data suggest that T cell response precedes humoral immunity and that the response to vaccination of CD4⁺ but not CD8⁺ T cells is associated to antibody level and improves with time after transplantation.