A red blood cell–based antigen delivery system to facilitate T cell epitope presentation to promote peripheral tolerance to ADAMTS13 in immune-mediated thrombotic thrombocytopenic purpura

Abstract

Background

Immune-mediated thrombotic thrombocytopenic purpura (iTTP) is a rare and potentially fatal autoimmune disease caused by a severe ADAMTS13 functional deficiency, mediated by autoantibodies targeting ADAMTS13. Despite high survival rates achieved with current treatments, between 20% and 50% of patients experience relapses. Ensuring longer-lasting recovery by restoring immune tolerance toward ADAMTS13 remains a significant unmet need.

Objectives

To address this, we aimed to exploit the tolerogenic nature of the red blood cell (RBC) clearance pathway by using RBCs as antigen carriers. This approach allows for persistent exposure to ADAMTS13–derived peptides, promoting antigen-specific attenuation of autoreactive CD4⁺ T cells and induction of regulatory T cells.

Methods

Specifically, we designed a fusion peptide comprising the trans-activator of transcription cell–penetrating peptide and an immunodominant ADAMTS13–derived T cell epitope (FINVAPHAR core amino acid sequence) to enable RBC surface loading.

Results

We found that the fusion peptide binds to RBCs in a concentration-dependent manner. To assess whether our approach supports antigen presentation via human leukocyte antigen (HLA) class II molecules, macrophages were incubated with peptide-loaded RBCs, followed by analysis of HLA-DR–bound peptides via mass spectrometry. FINVAPHAR-containing peptides of varying lengths were mainly presented on macrophages from donors carrying the iTTP-associated HLA risk allele DRB1∗11, confirming functional antigen presentation and supporting the biological relevance of our model. Hemoglobin-derived peptides were also abundantly presented, supporting erythrophagocytosis as the entry route of our trans-activator of transcription-FINVAPHAR fusion peptide.

Conclusion

Based on our results, we propose RBC-mediated delivery of ADAMTS13–derived peptides as a potential strategy for promoting tolerance in patients with iTTP.

Essentials

• •Relapses and long-term complications in iTTP persist.
• •A carrier peptide allows binding of an ADAMTS13–derived T cell epitope to RBCs.
• •Macrophages internalized peptide-loaded RBCs and presented ADAMTS13–derived peptides.
• •ADAMTS13 peptide–loaded RBCs may help restore tolerance and prevent relapse in iTTP.

1. Introduction

Immune-mediated thrombotic thrombocytopenic purpura (iTTP) is a rare and potentially fatal autoimmune disease caused by a severe ADAMTS13 functional deficiency mediated by autoantibodies targeting ADAMTS13 [1,2]. These antibodies accelerate clearance of ADAMTS13 from the circulation or decrease its proteolytic activity, resulting in insufficient cleavage of its substrate, ultra-large von Willebrand factor multimers. Accumulation of ultra-large von Willebrand factor strings on the endothelial surface promotes platelet interaction and formation of microvascular thrombi, leading to hemolytic anemia, thrombocytopenia, and organ failure due to reduced perfusion. The severity of the clinical presentation requires prompt diagnosis and intervention, as untreated iTTP is nearly universally lethal. The current standard of care includes therapeutic plasma exchange, corticosteroids, rituximab, and caplacizumab, which has greatly reduced mortality [3,4]. However, 20% to 50% of patients experience relapses [[5], [6], [7], [8]], highlighting the need for therapies that ensure long-lasting recovery by restoring immune tolerance.

Although the exact mechanisms leading to anti–ADAMTS13 autoantibody development remain unclear, a link with infections and microbiota through molecular mimicry has been proposed [9]. Autoreactive CD4⁺ T cells that escaped thymic negative selection may become activated when antigen-presenting cells (APCs) present pathogen-derived peptides that resemble ADAMTS13 epitopes. Alternatively, T cell activation may result from ADAMTS13–derived peptide presentation under pro-inflammatory conditions. Activated T cells further promote the autoimmune process by inducing autoreactive B cell differentiation into antibody-producing plasma cells. This process was confirmed by evidence for isotype switching and affinity maturation in anti–ADAMTS13 antibodies [[10], [11], [12], [13]]. CD4⁺ T cell involvement is further supported by the strong human leukocyte antigen (HLA) class II association in iTTP [[14], [15], [16], [17], [18], [19]] combined with ADAMTS13–derived peptide presentation on risk alleles [[20], [21], [22]]. A prime example of the latter is a peptide with the core sequence FINVAPHAR derived from the CUB2 domain of ADAMTS13. This epitope was found to be efficiently presented on major histocompatibility complex (MHC) class II molecules of dendritic cells derived from healthy HLA-DRB1∗11-positive donors [20,21] and was recognized by CD4⁺ T cells from HLA-DRB1∗11-positive TTP patients [22]. Notably, the DRB1∗11 allele is a risk factor for iTTP development in European Caucasian populations [[15], [16], [17]]. Moreover, the FINVAPHAR peptide demonstrated binding affinity to multiple HLA-DR molecules, suggesting promiscuity in its MHC class II presentation [20]. Taken together, these immunologic features demonstrate a central role for CD4⁺ T cells in iTTP and suggest that selectively targeting autoreactive T cells represents a promising therapeutic strategy. To achieve this, their immune tolerance to ADAMTS13 must be restored.

Apoptotic cell clearance is one of the most efficient physiological pathways for tolerance induction. Phagocytosis of apoptotic cells promotes an anti-inflammatory or suppressive phenotype in APCs and leads to presentation of autoantigens in a tolerogenic context [[23], [24], [25]].

Aging or damaged red blood cells (RBCs) are similarly cleared from the circulation by splenic and hepatic phagocytic cells in a process called erythrophagocytosis [26]. Building on this principle, various approaches have explored the use of RBCs as antigen-delivery vehicles for tolerance induction, including sortase-mediated RBC surface modification, erythrocyte-binding peptides, antibody-binding fragments, and microfluidics [[27], [28], [29], [30]]. These approaches resulted in protective effects by attenuating pathogenic T cell effector functions. Our group previously used a fusion peptide system using the transactivator of transcription (TAT) cell–penetrating peptide to anchor factor (F)VIII autoantigens on RBCs [31]. The TAT peptide, derived from the TAT protein of HIV, is known for its ability to cross the membrane of various cell types [32,33] and deliver molecular cargo for therapeutic and research purposes [34]. We demonstrated that TAT–FVIII RBCs were phagocytosed by macrophages, leading to FVIII presentation and T cell activation.

Building on our previous work, we investigated in this study whether RBCs can be used to deliver the ADAMTS13–derived immunodominant epitope FINVAPHAR to macrophages for MHC class II presentation, with the ultimate goal of inducing tolerance in iTTP.

2. Materials and Methods

2.1. Peptide design and synthesis

Peptides were synthesized by Biosynth BV, with a purity of >95%. The peptides synthesized included the TAT-CUB2 peptide (YGRKKRRQRRRGGCRLFINVAPHARIAIH) with (TAT-CUB2-biotin; molecular weight [MW] = 3954.8 Da) or without a biotin label (TAT-CUB2; MW = 3485 Da) and the CUB2 control peptide (CRLFINVAPHARIAIH) with (CUB2-biotin; MW = 2298.9 Da) or without a biotin label (CUB2; MW = 1830 Da).

2.2. Peripheral blood samples

RBCs and macrophages for the experiments outlined further were collected from whole blood samples from healthy volunteers. Fresh blood was drawn in accordance with Dutch regulations and following approval from the Sanquin Ethical Advisory Board and the Declaration of Helsinki. RBCs were isolated from plasma and buffy coat by centrifugation and washed twice in a solution of saline, adenine, glucose, and mannitol (Compoflex SAGM; Fresenius Kabi). RBCs were then washed twice in a HEPES-buffered saline solution (20 mM HEPES, 132 mM NaCl, 6 mM KCl, 1 mM MgSO₄, and 1.2 mM K₂HPO₄, pH 7.4) supplemented with 0.5% human serum albumin (HSA; Sanquin Plasma Products BV) for further experiments. Peripheral blood mononuclear cells were separated from other blood components by Ficoll-Paque PLUS (Cytiva) density gradient centrifugation, after which monocytes were isolated using anti-CD14 MicroBeads (Miltenyi Biotec), following the manufacturer instructions.

2.3. Flow cytometry

The binding of TAT-CUB2 peptides to RBCs was monitored by flow cytometry and imaging flow cytometry (Cytek ImageStream Mark II). RBCs (5 × 10⁶ per condition) were collected in HEPES-buffered saline supplemented with 0.5% HSA (Sanquin Plasma Products BV). To characterize peptide binding at different doses, the RBCs were incubated with increasing concentrations of TAT-CUB2-biotin peptide or CUB2-biotin control peptide (10, 20, or 30 μM) for 45 minutes at 37 °C. One condition included RBCs not incubated with any peptide, as negative control. Unbound peptides were removed using HEPES-buffered saline supplemented with 0.5% HSA. The cells were fixed in 0.1% glutaraldehyde (Sigma) for 20 minutes, and a subset was additionally permeabilized with 0.05% Nonidet P-40 Substitute (Sigma) for 10 minutes. Cells were washed and the biotin tag of bound peptides was stained with phycoerythrin (PE)-cyanine 7 (Cy7)–conjugated streptavidin (eBioscience; Invitrogen) and Alexa Fluor 488 (AF488)-conjugated streptavidin (Invitrogen). PE-conjugated anti-human glycophorin A (R&D Systems) was used to stain the RBC membrane. After washing, the RBCs were resuspended in 100 μL HEPES-buffered saline supplemented with 0.5% HSA and analyzed using a BD FACSCanto II (BD Biosciences). Flow cytometry data were processed using FlowJo, version 10.7.0.

For image analysis, 5 × 10⁶ RBCs were incubated with 20 μM peptide and fixed and stained as outlined earlier. Images were acquired on an ImageStream Mark II imaging flow cytometer (Cytek) and analyzed on IDEAS 6.2.23 software.

2.4. Macrophage differentiation and erythrophagocytosis of peptide-loaded RBCs

Monocytes were isolated from peripheral blood mononuclear cells as described earlier. For differentiation to macrophages, the monocytes were cultured in the presence of 0.05 μg/mL recombinant human macrophage colony-stimulating factor (PeproTech) in Iscove’s Modified Dulbecco’s Medium (Gibco) supplemented with 10% fetal calf serum (Bodinco), 2 mM l-glutamine (G7513; Sigma), 100 IU/mL penicillin, and 100 μg/mL streptomycin (P4333; Sigma). Differentiation continued for 10 days, with macrophage colony-stimulating factor replenished on days 3 and 7. Macrophages were harvested and 6 × 10⁶ cells per donor were incubated with 60 × 10⁶ peptide-loaded RBCs (loaded with 20 μM TAT-CUB2 peptide as described earlier) for 24 hours to allow phagocytosis.

2.5. HLA-DR peptide enrichment and sample preparation for mass spectrometry

Following incubation, cells were detached using a cell scraper and washed twice with Iscove’s Modified Dulbecco’s Medium. Cells were lysed in 1 mL of lysis buffer containing 10 mM Tris-HCl (pH 8), 0.25% n-octyl β-d-glucopyranoside (Sigma), 1% sodium deoxycholate (bioPLUS), and 10 mM EDTA (UltraPure 0.5 M EDTA pH 8; Invitrogen) and incubated end-over-end for 30 minutes at 4 °C. Cell lysates were cleared by centrifugation at 18,000g for 20 minutes at 4 °C, and supernatants were taken for subsequent immunoprecipitation.

HLA-DR–peptide complexes were isolated using CNBr-activated SepFast MAG 4HF magnetic beads (BioToolomics) coupled to anti-human HLA-DR antibody L243 (InVivoMAb; Bio X Cell). To control for nonspecific binding, an isotype-matched control antibody (C1.18.4 mouse IgG2a; InVivoMAb; Bio X Cell) was used. Immunoaffinity columns were prepared by coupling 2 mg of antibody to 1 mL of bead matrix. The cleared cell lysate of each donor was divided equally among 6 tubes containing ∼300 μL beads, with 3 containing L243-coupled beads and 3 containing isotype control beads. Samples were incubated overnight at 4 °C with continuous end-over-end rotation. Beads were washed twice with lysis buffer and 5 times with 10 mM Tris-HCl. Peptides were eluted from HLA-DR molecules by adding 500 μL of 10% acetic acid (Acetic Acid 100% for LC-MS LiChropur; Merck) on the column for 10 minutes at room temperature. The supernatant was collected from the beads and incubated at 70 °C for 10 minutes.

2.6. Mass spectrometry analysis

The peptide mixtures were loaded onto Evotip PureTips (Evosep) according to manufacturing guidelines. Gradient was set to a throughput of 15 samples per day using the manufacturer’s appropriate 15 × 150 μm, 1.5 μm, EV1137 Performance column (Evosep). For separation, 0.1% formic acid in water (Biosolve) and 0.1% formic acid in acetonitrile (Biosolve) were used as mobile phase A and B, respectively. Data were acquired on a timsTOF HT Mass Spectrometer (Bruker Daltonics) operated in ddaPASEF mode. The default polygon ion cloud filter, with a 1.17-second cycle time including 10 PASEF ramps, was used to select precursors for fragmentation. Accumulation time was set to 100 milliseconds with a 100% duty cycle and target intensity of 20,000 (threshold = 2500). A linear function of 1/k₀ (0.6 1/k₀ = 20 eV; 1.6 1/k₀ = 59 eV) was set for the collision energy and ion mobility was set from 0.6 to 1.6 1/k₀.

2.7. HLA-DR peptide search and identification

The raw mass spectrometry (MS) data files were analyzed using MSFragger (v3.8) in the FragPipe (v20.0) pipeline based on the settings as described in Phulphagar et al. [35]. Peptides and proteins were detected by querying the filtered human Swissprot database (release 2021.22.04) with added decoys (40,887 entries, 50% decoys). Peak matching precursor mass tolerance was set at −20 to 20 ppm, with a fragment mass tolerance of 10 ppm. Protein digestion was set to nonspecific with a peptide length of 8 to 30 amino acids and a peptide mass range 600 to 3000. Peptides were quantified using IonQuant (v1.9.8) using default settings without match between runs.

Output data were analyzed using R 4.2.3 (R Core Team) and Rstudio (2022.07.02) (Posit Team). Contaminants were filtered, label-free quantification (LFQ) values were log₁₀ transformed, and detected peptides were considered valid if quantified (LFQ value assigned) and present in 2 or 3 replicates per donor sample. Only peptides that were not detected in the isotype controls were considered as HLA-presented peptides. Proteins corresponding to these peptides were log₂ tranformed and analyzed.

3. Results

3.1. TAT-CUB2 peptides bind RBCs in a concentration-dependent manner

A peptide from the ADAMTS13 CUB2 domain containing the FIVNAPHAR core, with established relevance in iTTP and in DRB1∗11-positive patients in particular, was fused to the TAT sequence via a short glycine–glycine linker (GG) to generate a TAT-CUB2 peptide (Figure 1A). A CUB2 control peptide lacking the TAT sequence was used to assess non-specific binding. A biotin tag was added to allow detection of the fusion peptide in flow cytometry experiments.

Figure 1.

Monitoring of peptide binding to red blood cells (RBCs). (A) Schematic representation of the peptides used in this study. The TAT-CUB2 peptide consists of a cell-penetrating TAT peptide (dark blue), a short glycine linker (gray), and a CUB2-derived sequence (teal), with the major histocompatibility complex class II–presented core peptide underlined. The CUB2 control peptide lacks the TAT sequence. Peptides used for flow cytometry also include a biotin label for detection (between brackets). (B) Detection of peptide binding to RBCs via flow cytometry. RBCs were incubated with 10, 20, or 30 μM TAT-CUB2-biotin (blue) or CUB2-biotin control peptides (red). RBCs that were not incubated with peptide but otherwise followed the same procedure are represented in gray. Streptavidin–phycoerythrin (PE)–cyanine 7 (Cy7) fluorescence intensity (biotin detection) is represented on the x-axis, and the cell count (number of events) is represented on the y-axis. (C) Detection of peptide binding to RBCs via ImageStream. Representative brightfield and fluorescence images of RBCs after incubation with 20 μM TAT-CUB2-biotin (top) or CUB2-biotin control peptide (bottom). Peptides were visualized by their biotin label via streptavidin coupled to AF488 (green) and PE-Cy7 (pink); the membrane protein glycophorin A is stained by PE-conjugated anti-glycophorin A (orange). TAT, transactivator of transcription.

To determine whether the fusion peptide could bind to RBCs, RBCs were incubated with 10, 20, and 30 μM of TAT-CUB2 or CUB2 control peptide (Figure 1B). Binding was assessed by flow cytometry, using PE-Cy7-conjugated streptavidin for peptide detection via their biotin label and glycophorin A staining for RBC identification. RBCs incubated with TAT-CUB2 demonstrated a clear concentration-dependent increase in PE-Cy7 signal, with a pronounced rightward shift in histograms compared with the negative control and CUB2 control peptide, indicating the TAT-CUB2 peptide can bind RBCs efficiently (Figure 1B, top). Weak staining was also observed for the CUB2 control peptide, but the minimal signal increase suggests limited or nonspecific interactions.

To further investigate peptide localization, RBCs were permeabilized before staining to allow staining of peptides inside the cell. Histograms showed a similar profile when compared with non-permeabilized samples, suggesting TAT-CUB2 remains predominantly surface bound (Figure 1B, bottom).

ImageStream analysis confirmed that TAT-CUB2 peptides were surface bound (Figure 1C). In RBCs incubated with TAT-CUB2 peptide, AF488 (green) and PE-Cy7 (pink) signals show colocalization with the RBC membrane protein glycophorin A (PE stained, orange). AF488 and PE-Cy7 signals were not detected on RBCs incubated with CUB2 control peptide, further validating the importance of the TAT sequence for efficient binding to RBCs.

3.2. MS-based immunopeptidome analysis confirms FINVAPHAR presentation on HLA-DR

Our approach depends on efficient uptake of peptide-loaded RBCs by APCs, followed by proteolytic processing and subsequent presentation of the FINVAPHAR epitope on HLA class II molecules. This process is crucial for recognition of the epitope by ADAMTS13–specific CD4⁺ T cells via their T cell receptor (TCR). To investigate peptide presentation, we used primary monocyte–derived macrophages from healthy donors to ensure the experimental system closely resembles physiological antigen processing and presentation. Since our approach is primarily tailored to HLA-DRB1∗11, macrophages from 3 donors expressing this allele were used to investigate their ability to process and present FINVAPHAR on HLA class II molecules after erythrophagocytosis. We also included macrophages from non-DRB1∗11 donors to explore FINVAPHAR presentation by other HLA types and potential broader applicability of the peptide design (Figure 2A).

Figure 2.

Analysis of the DR-presented peptides after erythrophagocytosis. (A) Overview of HLA-DRB1 genotypes of donors used in this study. (B) Schematic overview of cell preparation, human leukocyte antigen (HLA) immunoprecipitation, and mass spectrometry workflow. (C) Heatmap representation based on label-free quantification (LFQ) intensity of detected peptides from HLA immunoprecipitates mapped to the TAT-CUB2 peptide. Peptide abundance is indicated by a red gradient. White indicates the peptide was not quantified. (D) Protein abundance rank based on log₂-transformed LFQ intensities. Names of the 5 most abundant proteins are given. Red blood cell–derived proteins and ADAMTS13 are highlighted in red and teal, respectively.

We developed and applied a MS-based immunopeptidomics workflow for analysis of HLA-DR–presented peptides after phagocytosis of TAT-CUB2 peptide–decorated RBCs by macrophages (Figure 2B). A negative control immunoprecipitation (isotype control–conjugated beads) was included to account for background signal. Each immunoprecipitation was performed in triplicate to account for technical variability.

Across all donors, the number of peptides eluted from HLA-DR molecules ranged between 5933 and 7179 (Supplementary Figure 1A). Peptides could also be detected for the isotype control immunoprecipitations (between 1050 and 1922). Conforming to established MHC class II–binding preferences, the identified HLA-DR–bound peptides had a normal length distribution centered around 15 to 16 amino acids [[36], [37], [38]] (Supplementary Figure 1B). In contrast, peptides of the isotype control exhibit a broader and more uniform length distribution, consistent with random background binding rather than specific HLA-DR presentation.

To visualize the variation between samples, we performed a principal component (PC) analysis based on the LFQ intensities of the immunopeptides (Supplementary Figure 1C). PC1 demonstrates donor-specific separation, reflecting individual differences in biological background, antigen-processing machinery, and protein expression patterns. Separation of HLA-DR immunoprecipitates from isotype controls across PC2 and PC3 shows discrimination between specific and background peptides. PC2 and PC3 also separate samples according to HLA types, with donor 4 clustering alongside DRB1∗11-positive donors (donors 1, 2, and 3), driven by their shared HLA alleles. The latter separation reflects the established impact of HLA polymorphisms on peptide-binding specificities.

For further analysis, peptides quantified in the isotype control were removed. Next, peptide presentation characteristics were analyzed by visualizing the overlap across donor peptidomes in an UpSet plot (Supplementary Figure 1D). The majority of the identified peptides were uniquely identified in donors 5 and 6 (5012 and 4370 peptides, respectively, accounting for 25% and 22% of identified peptides), consistent with their distinct HLA-DR profiles. The largest intersection of shared peptides was observed for donors 1, 2, and 3 (2231 peptides [11%]), as expected given their identical HLA type. These donors also exhibit a notable overlap with donor 4 (1779 peptides [9%]), likely due to the common DRB1∗15 allele. Additionally, 10% of all identified peptides (2001 peptides) were uniquely identified in donor 4, reflecting the contribution of the DRB1∗08 allele. The remaining 23% of peptides were distributed across smaller intersections. Overall, these findings are consistent with the donor information and known HLA II–binding preferences and support the reliability and biological relevance of the results.

We next examined peptides derived from the TAT-CUB2 peptide and compared their abundances across donors (Figure 2C). In total, 18 unique FINVAPHAR variants with variable lengths were detected, a characteristic result of N- and C-terminal trimming by endosomal or lysosomal proteases [[36], [37], [38]]. Presentation was particularly robust for DRB1∗11-positive donors 1, 2, and 3, as multiple variants with high LFQ values were detected. In addition, peptides containing FINVAPHAR could also be eluted from HLA-DR molecules of donor 4, albeit at lower LFQ intensities. Donors 5 and 6 seemed to only specifically support presentation of a RLFINVAPHARIAIH peptide, suggesting DRB1∗03:01 and DRB1∗07 molecules are less suitable for FINVAPHAR presentation.

A small subset of FINVAPHAR-containing peptides was also detected in the isotype controls (Supplementary Figure 1E), albeit generally with a lower LFQ value than HLA-DR immunoprecipitates (Supplementary Figure 1F). This finding underscores the importance of isotype controls to distinguish between HLA-presented peptides and background noise.

Finally, we investigated the source proteins from the identified peptides. To this end, we ranked the source proteins for each donor by their LFQ intensities (Figure 2D). Across all donors, the highest-ranking proteins originated from intracellular self-proteins involved in lysosomal functions (eg, cathepsins and NPC2), cytoskeletal structure (eg, FLNA and TUBA4A), and nuclear processes (eg, H3C1), a result of constitutive autophagy and lysosomal degradation feeding into the endogenous MHC-II presentation pathway [36]. Some donor-specific variation was observed (eg, CD4 and ITGB2 in donor 5 and ATP6V1A in donor 6), likely reflecting individual differences in biological background, HLA type, or antigen presentation efficiency. The detection of FINVAPHAR-derived peptides resulted in a ranking of the ADAMTS13 protein that was relatively high for donors 1, 2 and 3, while near the median for donors 4, 5 and 6. RBC-specific markers HBA2 and HBB (hemoglobin subunits) were among the most abundant proteins (average rank of 30 and 54 of 830 proteins, respectively), meaning RBC-derived antigens were also efficiently processed and demonstrating erythrophagocytosis as the TAT-CUB2 entry route.

4. Discussion

Although major advances in the diagnosis and treatment of acute iTTP have reduced mortality from >90% to <5%, relapses remain common and their prevention often requires ADAMTS13 monitoring and broad immunosuppression through preemptive rituximab [[5], [6], [7], [8]]. This persistent threat of relapse also contributes to substantial psychological burden on patients with iTTP [39]. In addition, iTTP survivors experience higher long-term mortality and morbidity, particularly due to cardiovascular and cerebrovascular complications [40,41]. Incomplete recovery of ADAMTS13 activity has emerged as a potential risk factor, suggesting that achieving higher ADAMTS13 activity levels during clinical remission would reduce these long-term complications [40]. Together, these challenges illustrate that future efforts should focus on understanding and improving long-term iTTP outcomes. Antigen-specific therapies to restore tolerance to ADAMTS13 may complement current therapies and contribute to this goal, by preventing relapse and reducing long-term complications.

In this study, we report the successful development of an RBC-based delivery system to facilitate presentation of an ADAMTS13–derived immunodominant epitope (FINVAPHAR) on MHC class II molecules of macrophages for induction of tolerance in iTTP. Loading of RBCs with a TAT-CUB2 fusion peptide allows anchoring of the FINVAPHAR epitope to the RBC surface and enables delivery to monocyte-derived macrophages via erythrophagocytosis, for subsequent processing and presentation via the endogenous MHC-II pathway. This strategy leverages the immune-silent nature of RBC clearance to redirect autoreactive CD4⁺ T cells toward anergic or regulatory phenotypes. Additionally, RBCs offer specific advantages as billions of RBCs are cleared daily after a long circulatory lifespan (∼120 days) [26], and we hypothesize that their abundance and persistence in the blood stream can extend and strengthen the tolerogenic response. Given that blood transfusions are one of the most commonly performed medical procedures, RBCs also represent a safe and biocompatible delivery vehicle.

Mechanistically, we observed that TAT-CUB2 associated with the RBC membrane in a concentration-dependent manner and remains surface-bound rather than internalized. This effect is likely explained by the hydrophobic nature of the FINVAPHAR sequence: the high content of amino acids with hydrophobic side chains (F, I, V, A, and P) likely restrict the peptide from migrating past the lipid bilayer. As surface anchoring preserves effective antigen uptake upon erythrophagocytosis, this does not negatively impact our approach.

To address the inherent challenges of immunopeptidomics (such as low abundance of individual HLA-presented peptides, limited recovery, and absence of predefined cleavage sites), we developed a high-throughput, high-sensitivity MS workflow using trapped ion mobility time-of-flight MS (Bruker timsTOF) in combination with MSFragger software for data analysis. Stringent search parameters were set to match physicochemical properties of naturally processed MHC class II–presented peptides. High peptide yield, consistency across technical replicates, and clear HLA-DR/isotype separation confirm the robustness and sensitivity of our protocol. By integrating LFQ-based quantification, isotype controls and peptidome validation based on HLA-DR–matched overlap, we deliver analytical resolution that is similar to dedicated immunopeptidomics studies [35] and that was vital to reliably detect FINVAPHAR presentation among thousands of HLA-DR–presented peptides.

Peptide presentation was most prominent on macrophages from DRB1∗11-positive donors, consistent with previous studies describing highly efficient FINVAPHAR presentation by this allele [20,21]. Macrophages from donor 4 with HLA type DRB1∗08/∗15 also supported presentation of FINVAPHAR-derived peptides, indicating overlapping binding capabilities with DRB1∗11. We lack data to claim whether the observed presentation in donor 4 can be attributed to DRB1∗08 or DRB∗15 molecules or both, but peptide-binding specificity analysis of HLA-DR alleles has suggested similar binding profiles for DRB1∗08 and DRB1∗11 [42,43]. Interestingly, DRB1∗03:01 and DRB1∗07 molecules (HLA alleles of donors 5 and 6, respectively) presented only a single peptide variant, indicating these HLA molecules are not that permissive for FINVAPHAR presentation. Nevertheless, the most robust presentation was observed in DRB1∗11-positive individuals, and this is especially relevant for iTTP as DRB1∗11 is a known risk allele in European Caucasian populations. Approximately 62% of French and 48% of German patients with iTTP test positive for the DRB1∗11 allele, an overrepresentation compared with their reference populations [15,17]. While the reported risk allele in Japanese iTTP is DRB1∗08:03, its predicted binding motif (LFINVAPHA) shifts only 1 amino acid compared with FINVAPHAR [19], suggesting that the current TAT-CUB2 design may also be relevant in this population. The higher incidence of iTTP in Black people is explained by a lower frequency of the protective DRB1∗04 allele rather than the presence of a clearly defined HLA risk allele [44], complicating the use of antigen-specific approaches such as ours in this population. However, the modularity of our approach allows substitution for alternative dominant epitopes, enabling adaptation for other patient populations and even other autoimmune diseases.

The use of primary monocyte–derived macrophages permitted antigen processing in a physiologically relevant manner, resulting in a collection of trimmed N- and C-terminal variants of the same core sequence [[36], [37], [38]]. As peptide variants can differ in TCR affinity and peptide–MHC stability, the range of autoreactive T cells that can be targeted is broadened [45,46]. Peptide presentation following RBC uptake occurs in the absence of costimulatory signals [47], creating an environment in which even pre-existing effector T cells can be reprogrammed into anergic or regulatory phenotypes due to their intrinsic plasticity [[27], [28], [29], [30],48,49]. Importantly, robust peptide presentation was observed already after 24 hours of coincubation, while RBCs have an average in vivo lifespan of 120 days. In a related study on RBC-induced tolerance, deletion of antigen-specific T cells in mice occurred within 3 days when 90% of transfused RBCs still remained [27], demonstrating the potential durability and rapid onset of tolerance using this approach.

The current study has several potential limitations that need consideration. First, we have not yet been able to functionally validate peptide presentation, due to the scarcity of patient-derived autoreactive T cells, lack of T cell hybridomas carrying FINVAPHAR-specific TCRs and the absence of a fully representative mouse model that reproduces human iTTP [50]. Nonetheless, the detection of FINVAPHAR peptides on MHC class II molecules is a prerequisite for direct T cell engagement and supports the feasibility of this strategy. Second, although immunogenicity to the minimal TAT cell-penetrating sequence (YGRKKRRQRRR) when used for its delivery properties is not described, we cannot exclude immune responses to the TAT sequence or potential neo-epitopes created by fusing TAT to the CUB2 peptide. While the tolerogenic mechanism of action of our platform suggests potential immunogenicity toward the TAT peptide is avoided, it is important that potential immune responses against RBCs modified with TAT-CUB2 peptide are carefully evaluated in follow-up studies. Third, as HIV is recognized as a precipitant for iTTP [51], HIV-positive patients with iTTP represent a distinct subgroup in whom the use of an HIV-derived peptide fragment may have specific implications. Despite evidence suggesting that immune responses to TAT are generally weak and occur only in a minority of HIV-infected individuals [[52], [53], [54], [55], [56], [57]], special attention should be paid to HIV-positive patients with iTTP in subsequent preclinical studies to identify potential interactions with HIV-specific immunity.

Our findings support the potential of RBC-mediated peptide delivery to promote antigen-specific tolerance in iTTP. We propose administering this therapy during remission, when the patient is clinically stable, ADAMTS13 activity is restored and systemic inflammation is no longer observed. Even though immunological considerations suggest the remission period as the optimal moment, it should be empirically determined in practice which timing maximizes successful tolerance induction.

While further functional and in vivo studies are needed, the modular platform presented in this study holds promise for sustained immune reprogramming in iTTP and possibly other autoimmune diseases.

Supplementary material

Supplementary figure 1

Profiling of the HLA-DR immunopeptidome for validation of reliability. (A) The number of unique peptides identified per replicate in HLA and isotype conditions. (B) Peptide length distribution of HLA-DR-presented peptides and isotype control peptides. (C) Principal component (PC) analysis of peptidomes across PC1, PC2 and PC3. Donors are indicated by color, HLA and isotype conditions by circles and triangles, respectively. (D) UpSet plot comparison of identified peptides across all donors. Set size represents the total number of peptides identified per donor and intersection size represents the number of shared peptides identified in different donors. (E) Heatmap representation based on LFQ intensity of detected peptides from isotype controls mapped to the TAT-CUB2 peptide. Peptide abundance is indicated by a red gradient. White indicates the peptide was not quantified. Asterisks indicate peptides detailed in F. (F) Comparison of LFQ intensities for three TAT-CUB2-derived peptides between HLA and isotype conditions. Donors are indicated by color, solid lines show average LFQ intensity per donor, dotted lines connect the corresponding LFQ intensity averages of HLA and isotype conditions.