We collected blood samples by using vascular access before the start of dialysis or by venipuncture for the control population 16 weeks after the standard 2-dose vaccination with a 21-day interval of BNT162b2 (Pfizer-BioNTech, https://www.pfizer.com) was completed (T2). An analysis of samples from this population that were collected 3 weeks after the second dose of BNT162b2 (T1) has been published previously (13). A total of 76 patients on maintenance hemodialysis and 23 healthcare workers from the same dialysis center participated in the longitudinal follow-up (13). Demographic characteristics (e.g., age and sex), body mass index, time on dialysis, use of immunosuppressive medications, and anti-S1 domain IgG levels at T1 of persons who did not provide a sample at T2 were not substantially different compared with persons included in this analysis (Table; Appendix Table 1, 2). We obtained plasma by using an S-Monovette lithium heparin blood collection kit (Sarstedt, https://www.sarstedt.com). We used whole-blood samples immediately for an interferon-γ (IFN-γ) release assay (IGRA). To inactivate potential pathogens, we treated collected saliva samples with Tri (n-butyl) phosphate for a final concentration of 0.3% and Triton X-100 for a final concentration of 1%.

The study was approved by the Internal Review Board of Hannover Medical School (approval number 8973_BO-K_2020). We obtained written informed consent from all participants before the start of the study.

We coupled antigens to spectrally distinct MagPlex beads (Luminex, https://www.luminexcorp.com) by using EDC/s-NHS coupling for all standard (MULTICOV-AB) antigens (14). We coupled receptor-binding domains (RBDs) from VOCs by using Anteo coupling (AnteoTech, https://www.anteotech.com) according to the manufacturer’s instructions (15).

We analyzed IgG and IgA binding and levels by using MULTICOV-AB, a multiplex coronavirus immunoassay, as previously described (14). For our study, we used a panel of recombinant proteins as antigens (Appendix Table 3). In brief, we immobilized antigens on spectrally distinct populations of MagPlex beads either by EDC/s-NHS coupling (14) or by Anteo coupling according to the manufacturer instructions (15). We then incubated the combined MagPlex beads with samples. After conducting a wash step to remove unbound antibodies, we detected IgG or IgA with either R-phycoerythrin labeled goat anti-human IgG (Jackson ImmunoResearch, https://www.jacksonimmuno.com) or IgA (Jackson ImmunoResearch) as secondary antibodies. After conducting another wash step and bead resuspension, we measured samples once on a FLEXMAP 3D instrument (Luminex) by using the following settings: timeout, 80 s; gate, 7,500–15,000; reporter gain, standard photomultiplier tube; 40 events. Raw median fluorescence intensity (MFI) values or normalized values (MFI/MFI of quality control [QC] samples) (15) are reported. Three QC samples were measured per individual plate to monitor MULTICOV-AB performance. We measured all samples once.

We carried out an angiotensin-converting enzyme 2 receptor-binding domain (ACE2-RBD) competition assay as previously described (15; D. Junker et al., unpub. data, https://doi.org/10.1101/2021.08.20.21262328) to determine IgG neutralization capacity against SARS-CoV-2 wild-type and the VOCs. For this assay, we combined biotinylated ACE2 with individual samples (and as a control, ACE2 alone) and incubated with the previously described MULTICOV-AB bead mix. Before and after ACE2 detection with Streptavidin-PE (Moss, Fisher Scientific, https://www.fishersci.com), we conducted washes. We measured samples once on a FLEXMAP 3D instrument with the same settings as MULTICOV-AB and analyzed them by using normalization of MFI values against the control. We considered samples with a neutralization ratio <0.2 as nonneutralizing. This cutoff is based on comparison to a classic virus neutralization test (D. Junker et al., unpub. data).

As a control for the MULTICOV-AB results, we also analyzed plasma samples by using the Anti-SARS-CoV-2 QuantiVac ELISA IgG (Euroimmun, https://www.euroimmun.com). Samples were measured as previously described (13). We measured all samples once.

We analyzed SARS-CoV-2–specific T-cell responses from whole blood by measuring IFN-γ production after stimulation with a peptide pool from the SARS-CoV-2 spike S1 with the SARS-CoV-2 Interferon Gamma Release Assay (Euroimmun) and the IFN-γ ELISA (Euroimmun), as previously described (13). We subtracted background signals from negative controls and calculated final results in milli-IU (mIU) per milliliter by using standard curves. Results from positive and negative controls were not statistically significantly different between timepoints T1 and T2. We considered IFN-γ concentrations >200 mIU/mL as reactive. We defined this arbitrary cutoff by using average background IFN-γ activity without antigen-stimulation in all samples of T1 multiplied with 10 for the threshold for IGRA positive. Using this cutoff, we found negative IGRA results in all of the 15 control samples (prepandemic persons) (16). The upper limit of reactivity was 2,000 mIU/mL.

We matched sample metadata and collecting results from different assay platforms in Microsoft Excel 2016 (https://www.microsoft.com). We used GraphPad Prism 8.4.3 (https://www.graphpad.com) for statistical analysis. We generated figures in RStudio 1.2.5001 running R 3.6.1 (https://www.rstudio.com). We used the beeswarm add-on package to visualize data as strip charts with overlaying boxplots and to create nonoverlaying datapoints and used the RcolorBrewer add-on to generate specific colors for plots. We then edited the figures by using Inkscape 0.92.4 (https://inkscape.org).

Because antibody levels are considered a proxy for protection, we initially examined the seroreversion rate by using MULTICOV-AB (14), a previously published bead-based multiplex immunoassay that simultaneously analyses >20 different SARS-CoV-2 antigens, including the RBDs of VOCs and the endemic human coronaviruses. Similar to findings from our previous report (13), RBD IgG responses within the dialysis group (median normalized MFI 4.26 among 76 patients) toward SARS-CoV-2 wild-type RBD were significantly reduced compared with those for the control group (median normalized MFI 13.6 among 23 persons; p<0.001) (Figure 1, panel A) 16 weeks after complete vaccination (T2). Compared with titer levels at 3 weeks after the second dose (T1), at 16 weeks after (T2), antibody titers had significantly decreased, by 61% in the control group and 75% in the dialysis group (p<0.001) (Figure 1, panel A). RBD IgG levels measured by MULTICOV-AB were additionally verified with a commercial quantitative in vitro diagnostic antibody test (Spearman rank 0.956) (Appendix Figure 1). Although none of the samples of the control group were classified as seronegative (titer below the cutoff) (Appendix Figure 2), 19.7% (15/76) of dialysis samples were defined as such 16 weeks after the second dose (T2), which constitutes a substantial increase from 3 weeks after second vaccination (T1), at which point only 5.3% (4/76) of samples were seronegative. When examining plasma titers against nucleocapsid, we did not observe any dialysis patients, other than one who had a PCR-confirmed infection before the first dose, having a value above the cutoff that would indicate infection.

To evaluate whether this reduction in plasma RBD IgG was also present at the mucosal site, we profiled the local antibody response in saliva by using MULTICOV-AB. As observed in plasma, a significant reduction occurred in saliva RBD IgG titers in the dialysis (median 143 among 71 patients) compared with the control group (median 313.5 among 23 persons) (p = 0.02) (Figure 1, panel B). When comparing saliva RBD IgG levels at T1 to those at T2, we observed a statistically significant decline in both groups (p<0.001) (Figure 1, panel B), suggesting that the antibodies have potentially lost competence to prevent transmission if infected. When examining RBD IgA, we observed a significant difference in titers between persons in the control and dialysis groups (p = 0.003) (Appendix Figure 3, panel A); 47.8% of controls and 75% of dialysis patients were classified as seronegative. This more pronounced reduction in IgA versus IgG levels most likely represents the shorter IgA half-life. Saliva RBD IgA tended to be higher in the dialysis group, although not significantly (p = 0.051) (Appendix Figure 3, panel B).

We next examined whether neutralization potential was also hindered because solid evidence exists on the protective role for neutralizing serum antibodies (17). By using an ACE2–RBD competition assay, which assesses neutralization potency toward SARS-CoV-2 wild-type and the circulating Alpha, Beta, Gamma, and Delta VOCs, we found that neutralization against wild-type SARS-CoV-2 RBD was significantly reduced in the dialysis group compared with the controls (p<0.001) (Figure 1, panel C) 16 weeks after complete vaccination. We found that 82.6% (19/23) of control samples and 89.5% (68/76) of dialysis patient samples were below the 0.2 threshold, which indicates the absence of neutralizing activity (Appendix Figure 2), a threshold is based on information provided for other available ACE2 competition assays (18). This difference represents a substantially significant reduction (p<0.001 for both groups) in neutralizing activity compared with 3 weeks after second vaccination, at which point only 4.3% (1/23) of the control samples and 50.0% (38/76) of the dialysis patient samples were below the threshold (Figure 1, panel C; Appendix Figure 2).

Because some persons might be able to control and clear SARS-CoV-2 infections with a strong T-cell response alone, we examined spike-specific SARS-CoV-2 T-cell responses by using a commercially available IGRA. Although absolute mean IFN-γ responses in the dialysis group compared with the control group tended to be lower (median 370 vs. 651 mIU/mL), this difference was not significant (p = 0.13) (Figure 1, panel D). In the control group, IFN-γ release after restimulation declined significantly from the first timepoint (median 1,505; p<0.001) (Figure 1, panel D), whereas for dialysis patients, this decline was not significant (median 580; p = 0.13) (Figure 1, panel D). This difference is probably attributable to most control samples being at the assay’s upper limit of detection at the first timepoint, when the dialysis samples already showed reduced IFN-γ release. Overall, the number of nonresponders was higher in the hemodialysis group (40.8% [31/76]) than the control group (21.7% [5/23]) (Appendix Figure 2). A lack of serologic response appears to be more driven by T-cell immunity than B-cell immunity; 2.6% (2/76) of the dialysis group having a T-cell response but no B-cell response, compared with 23.6% (18/76) who had a B-cell response but no T-cell response. In total, 17.1% (13/76) of the dialysis group were classified as complete nonresponders because of the absence of detectable SARS-CoV-2 wild-type B- and T-cell responses, compared with none in the control group.

Having characterized response against wild-type SARS-CoV-2, we then assessed humoral response against the VOCs Alpha, Beta, Gamma, and Delta. As shown with classical cell-culture based virus neutralization assays (7), neutralization responses were also reduced for all VOCs compared with wild-type when we used the previously described ACE2-RBD competition assay. Compared levels at with the initial timepoint, neutralization decreased significantly for the Alpha and Beta VOCs (p<0.001 for both) (Figure 2, panel A, B). We were unable to determine these changes for Gamma and Delta because these variants were not measured in the initial analysis. In a comparison between the dialysis and the control cohort, dialysis patients had significantly reduced neutralization against Alpha (p<0.001) (Figure 2, panel A), Gamma (p = 0.014) (Figure 2, panel C), and Delta (p = 0.002) (Figure 2, panel C) but not for Beta (p = 0.08) (Figure 2, panel B). The number of nonresponders was variable between the different strains although consistently high; 87.0% of the control group and 93.4% of the dialysis group were considered nonresponders against Alpha, 95.7% of the control group and 100% of the dialysis group against Beta and Gamma, and 95.7% of the control group and 96.1% of dialysis group against Delta.