Seven-week-old C57BL/6J female mice were purchased from The Jackson Laboratory (strain# 000664 Bar Harbor, ME, USA) and maintained under specific pathogen-free conditions. All animal studies were approved by the local Institutional Animal Care and Use Committee. ApoB_978–993 (P6: TGAYSNASSTESASY) peptide was purchased from A&A Labs (San Diego, CA). Complete (CFA) and incomplete (IFA) Freund’s adjuvants were purchased from SIGMA (St. Louis, MO). C57BL/6J mice were intramuscularly immunized with 100 μg P6 peptide in 100 μl CFA/PBS (50:50%) at week 0 and 100 μg P6 peptide in 100 μl IFA/PBS (50:50%) at week 2. For each immunization 50 μl were injected into the left and right musculus quadriceps femoris. Two weeks after final immunization the inguinal and para-aortic lymph nodes were collected.

ApoB:MHC monomers were expressed as previously described (11). Briefly, sequences encoding the antigenic peptide ApoB 978–993 were fused to the N-terminus of the mouse MHC-II (I-A^b) beta chain by a flexible polyglycine linker in the pRMHa-3 expression vector and co-expressed in Drosophila melanogaster S2 cells with the mouse MHC-II (I-A^b) alpha chain and BirA ligase. Biotinylated ApoB:MHC monomers were purified from culture supernatants using nickel affinity chromatography, followed by an additional purification on a Pierce Monomeric Avidin UltraLink Resin column (Thermo Fisher Scientific, Waltham, MA, USA) and coupled to streptavidin-phycoerythrin (PE) or streptavidin-allophycocyanin (APC) (Prozyme, Hayward, CA, USA) to generate tetramers.

Cell suspensions were prepared from draining (inguinal and para-aortic) lymph nodes. Lymph nodes were passed through a 70 μm cell strainer and the cell suspension was filtered once more through a 70 μm strainer before washing (400 × g, 4 min, room temperature). Cells were counted using trypan blue and a Neubauer chamber. Before incubation with tetramers, CD4 + T cells were enriched by a negative magnetic bead separation followed by anti-mouse biotinylated monoclonal antibodies all obtained from Tonbo Biosciences, San Diego, CA, USA:

CD8α, clone 53-6.7, cat.# 30-0081-U500; CD19 clone 1D3, cat.# 30-0193-U500; NK1.1, clone PK136, cat.# 30-5941-U500; CD11c, clone N418, cat.# 30-0114-U500; F4/80, clone BM8.1, cat.# 30-4801-U500; CD11b clone M1/70, cat.# 30-0112-U500; TER-119, clone TER-119, cat.# 30-5921-U500; CD45R, clone RA3-6B2, cat.# 30-0425-M001; and streptavidin-coupled magnetic microbeads (Stem Cell Technologies, Seattle, WA, USA). 20 μl of each antibody and 70 μl magnetic beads were used for each mouse.

Enriched CD4 + T cells were > 90% pure, based on surface expression of TCR-β and CD4 as measured by flow cytometry. Isolated CD4 T cells were washed (400 × g, 4 min at 4°C) and incubated for 30 min at 37°C and 5% CO₂ in 100 μl of 50 nM Dasatinib (Stem Cell Technologies) in RPMI supplemented with 10% FCS (Gemini Bio, West Sacramento, CA, USA) and 1x Pen/Strep (Thermo Fisher Scientific). Subsequently, 1 μl (equaling a final concentration of 10 nM) of ApoB:MHC-streptavidin-PE and ApoB:MHC-streptavidin-APC tetramer were added for an additional 45 min incubation period. Cells were washed (400 × g, 4 min at 4°C), the supernatant discarded and resuspended in 100 μl of Live/Dead Aqua (Thermo Scientific Fisher) diluted 1:1,000 in PBS. After a 30 min incubation on ice in the dark, 100 μl of staining buffer (fetal bovine serum 1:50 diluted in PBS) were added to the samples and cells were washed (400 × g, 4 min at 4°C). The supernatant was discarded and the cells were resuspended in staining buffer with the following anti-mouse monoclonal antibodies at 1:100 dilution:

CD16/32, clone 2.4G2, Tonbo Biosciences, cat.# 70-0161-M001; CD8a BV421, clone 53-6.7, BioLegend, San Diego, CA, USA, cat.# 100738; CD19 BV421, clone 6D5, BioLegend, cat.# 115538; NK1.1 BV421, clone PK136, BioLegend, cat.# 108732; CD11c BV421, clone N418, BioLegend, cat.# 117343; F4/80 BV421, clone BM8, BioLegend, cat.# 123132; CD11b BV421, clone M1/70, BioLegend, cat.# 101251; Ter119 BV421, clone TER-119, BioLegend, cat.# 116234, CD45R BV421, clone RA3-6B2, BioLegend, cat.# 103251; CD25 BV605, clone PC61, BioLegend, cat.# 102036; CD62L BV785, clone MEL-14, BioLegend, cat.# 104440; TCR-β Alexa Fluor 700, clone H57-597, BioLegend, cat.# 109224; CD4 APC-eFluor 780, clone GK1.5, Thermo Fisher Scientific, cat.#47-0041-82; CD44 PE/Cy7, clone IM7, BioLegend, cat.# 103030.

After staining for 30 min on ice in the dark, cells were washed (400 × g, 4 min at 4°C), the supernatant was discarded, and cells were resuspended in 150 μl staining buffer for sorting with a BD Aria Fusion (BD Biosciences, San Diego, CA, USA). A 70 μm nozzle and medium pressure at 800–1,200 events/s were used to index sort single cells into single wells of a 384-well plate (Thermo Fisher Scientific, 4483322, MicroAmp EnduraPlate optical). For sorting P6:MHC-positive CD4 T cells, 90% of the sample volume were used while P6:MHC-negative CD4 T cells were sorted from the remaining 10%. Each well contained 4 μl lysis buffer. For 1 reaction of lysis buffer 2 μl of diluted RNase inhibitor, 1 μl dNTPs (10 mM, Thermo Fisher Scientific) and 1 μl of 10 μM Oligo-dT30VN (IdT, Coralville, IA, USA) were mixed. RNase inhibitor (Takara/Clonetech, Mountain View, CA, USA) was diluted 1:20 in 0.2% Triton X-100 (Merck, Darmstadt, Germany), which was diluted in nuclease-free water (Qiagen, Hilden, Germany)

After sorting, the plate was sealed, gently vortexed, and centrifuged at 2,000 rpm for 30 s ensuring that all cells were collected in the lysis buffer. Subsequently, the plate was incubated for 3 min at 72°C to hybridize the Oligo-dT primer with the mRNA. The plate was immediately stored at -80°C until library preparation.

Single cell libraries were prepared according to the Smart-Seq2-protocol (22, 23) with the following modifications. Pre-amplification PCR cycles were increased to 23 to obtain sufficient amounts of cDNA for the sequencing analysis. Primer dimers were eliminated by two 0.8x Ampure-XP bead clean ups. 0.3–0.5 ng of pre-amplified cDNA were subjected to library preparation with the Nextera XT library preparation kit, Illumina, San Diego, CA, USA in 8 μl reaction volume. Barcoded libraries were pooled and sequenced the with a S1 flow cell and a 300-cycle kit on a NovaSeq Illumina platform to obtain 150-bp paired-end reads. Quality controls were performed using TapeStation with D5000 high sensitivity tapes (both from Agilent, Santa Clara, CA, USA). Quantification was performed using Qubit high sensitivity kit (Thermo Fisher Scientific) after PCR preamplification, and the PicoGreen assay (Thermo Fisher Scientific) after Nextera XT had been performed.

A total of 150 bp paired-end raw sequencing reads from Illumina NovaSeq (Illumina pipeline v1.9) for 176 cells were run through FASTQC¹ (v0.11.7). All cells passed sequence quality score as defined by FASTQC. On average, we captured 2.5 million reads per cell. STAR (v2.6.0a) (24) index was built using ENSEMBL GRCm38 genome (primary assembly) and annotation (GRCm38.92) (25). Raw reads were then mapped using STAR with the added parameters “–quantMode GeneCounts” and “–outSAMunmapped Within”. The mapped reads BAM file was sorted using SAMtools (v0.1.19) (26). Post-mapping sequencing quality was performed on the sorted BAM file using QoRTs (v1.3.0) (27). Gene body coverage was determined for the upper-middle quartile genes according to QoRTs. Mapping rate includes unique as well as multi-mapping reads.

Gene counts were obtained from STAR. ENSEMBL gene ids were converted to gene names using the ENSEMBL annotation (25). Genes which are not within chromosomes were discarded. Ensembl gene ids without a corresponding gene name were discarded. Identical gene names were collapsed using average. Seurat (28) was used to analyze the scRNA-Seq dataset. We discarded cells with less than 500,000 reads, or less than 2,000 genes, cells with high mitochondrial content (= 10%), or cells with low total mapping rate (=75%). All analyses were performed using the Seurat (v3.0.2) R package. The expression of the remaining 129 cells were normalized (LogNormalization with scale.factor = 1e6) and imputed using ALRA. The top 500 variable genes using the “vst” method were selected and scaled, regressing out the effects of mitochondrial content, plate, and sequencing depth. PCA was performed and the first 8 PCs according to JackStraw method were retained. UMAP (29) projection was done on these PCs, with number of neighbors set to 50. Clustering was done using Louvain algorithm, with resolution set to 0.3 and k.param for neighbor finding set to 10. Marker identification was done using the ROC test, retaining only up-regulated genes with an AUC = 0.8. g:Profiler (30) was used to find over-represented GO terms and pathways in the marker genes. Statistically significant pathways with a q-value of =0.1 are reported. Raw FASTQ files were fed into TraCeR (v0.6.0) (16) “assemble” to reconstruct the alpha and beta chains of the TCR. Default arguments were used. We used all the reconstructed chains to call TCR clonotypes regardless of productivity or expression level. Heatmaps were created using pheatmap (v1.0.12). Chain similarity graph was done using ggraph (v1.0.2).

The three I-A^b crystal structures PDB ID 3C60, 2IAD, and 1AIO were superimposed, revealing conserved backbone binding of the three different peptides in the I-A^b binding groove. We used NetMHCIIpan (v3.2) (31) to predict the binding core of the ApoB6 peptide, which was then superimposed with the binding cores of the three peptides from the available crystal structures. The ApoB6 peptide was manually built in COOT (32) to identify possible TCR contact residues.

We next superimposed five complex crystal structures with different I-A restricted TCR’s which demonstrated a common mode of binding above the I-A^b peptide binding groove. Of all TCRs, the β chain was more conserved in sequence compared to the α chain. The TCRs bound to I-A^b with similar footprint. Thus, we assumed that the β chain dictated the conserved binding orientation and, therefore, we modeled our β chain on top of the TCRβ chain that has the highest sequence identity. A multiple sequence alignment of these TCRs with the major clone (TRAV7D-6/TRAJ17, TRBV13-2, TRBJ2-4) identified in this study revealed an almost conserved sequence with the TCRβ chain of YAe62 (33) (PDB ID 3C60), with the exception of CDR3β. We modeled our TCR sequence by superimposition with the TCRβ chain of YAe62 and manually rebuilt the CDRs, where they differed in length and sequence, to identify possible contact residues with the ApoB6 peptide. The B3K506 (PDB ID 3C5Z) TCRβ chain was slightly more diverse in sequence and was excluded as a template for modeling.

We immunized C57BL/6J mice with ApoB P6 (ApoB978-993 TGAYSNASSTESASY) in CFA i.m. followed by P6 in IFA 2 weeks later and harvested the draining (inguinal and para-aortic) lymph nodes at 4 weeks after the primary immunization (Supplementary Figures 1A–C). Antigen-experienced CD4 T cells were identified as TCRβ⁺CD4⁺CD44^hiCD62L^–. All 32 P6:I-A^b tetramer-PE and -APC positive cells (P6⁺) from one ApoB-P6 immunized mouse and 144 tetramer-negative antigen-experienced CD4 T cells were sorted into single wells. Next, libraries were prepared (NexteraXT) and sequenced. All P6⁺ and 97 out of 144 P6^– cells passed quality filters (Supplementary Figures 2, 3). TraCeR was used for TCR assembly (16). In 23 out of 32 P6⁺ cells (71%) and in 53 out of 97 P6^– cells (54%), both TCRα and β chains were successfully reconstructed. Both productive (expressed) and non-productive rearrangements were used to call clonotypes (Figure 1 and Supplementary Table 1). Clonal expansion (one major, two minor clones) was identified in the P6⁺ cells. Although considered a different clonotype (based on productive and non-productive α and β chains), several other unexpanded cells shared the same α or β chain with that major clone (Figure 1). The most common ApoB-P6-specific TCRα and β chains contained TRAV7D-6 and TRAJ17 as well as TRBV13-2 and TRBJ2-4 (Figure 2 and Supplementary Table 1).

Clustering of cells according to their transcriptional profile revealed three clusters (Figure 3A). Although these transcriptional clusters were not clearly separated, overlaying the TCR clonotype information revealed good matching to groups composed mostly of expanded P6⁺ cells, non-expanded P6⁺ cells and P6^– cells (Figure 3B). Unlike the major expanded clone (10 cells; orange) and the moderately expanded clone (4 cells; green) which were in the expanded P6⁺ cluster, the minor expanded clone (2 cells; blue) was among the P6^– cluster (Figure 3B). Expression of Foxp3, the lineage-defining transcription factor (TF) of T_regs, and several other T_reg-related genes such as the co-inhibitory molecule cytotoxic T-lymphocyte associated protein 4 (Ctla4), the TF helios (Ikzf2), the tumor necrosis factor receptor superfamily member 18 (Tnfrsf18, CD379), or the capping actin protein (Capg) were upregulated in expanded compared to non-expanded P6⁺ cells (Supplementary Table 2). Gene Ontology (GO) enrichment analysis further revealed that the expanded P6⁺ cell marker genes were enriched for terms related to lymphocyte development, T cell activation, cellular response, transport and fiber organization (Supplementary Table 3).

We compiled a list of 39 T_reg, T helper 1 (T_H1), 2 (T_H2), 17 (T_H17), and T follicular helper cell (T_FH) lineage-defining genes that were highly expressed [imputed counts per million (cpm) =10 in at least three cells] to determine the lineage of each individual P6⁺ cell. Hierarchical clustering based on these genes showed four clearly distinguishable transcriptomic profiles (Figure 4). Most of the expanded P6⁺ cell transcriptomes were in clusters 1 and 2. Cells in cluster 2 highly expressed 11 of 12 T_reg genes including Foxp3, Ctla4, Ikzf2, Tnfrsf18, and Il2ra, the gene encoding the high affinity IL-2 receptor CD25. Additionally, a few T_H17 genes were expressed by the cells in cluster 2, most prominently Rora and, to a lesser extent, Rorc, which encodes the lineage-defining TF RORγT. The cells in cluster 2 also expressed some T_H2 genes, such as Gata3, which encodes the lineage-defining TF GATA3, and the TFs Batf and Runx3. In summary, this gene signature defines cluster 2 as expanded and effector T_regs (Figure 4). The transcriptomic profile of cluster 1 was similar to cluster 2, although the T_reg signature was weaker. Four P6⁺ cells formed cluster 3. These cells had a clear residual T_reg signature, but highly expressed T_H1-related genes such as the Tbx21, encoding the lineage-defining TF T-bet, interferon gamma receptor 1 (Ifngr1), or nuclear factor of activated T-cells, cytoplasmic 1 (Nfatc1). In addition, this cluster showed particularly high expression of T_reg marker genes that are also markers of T cell activation (Ctla4, Il2ra) (34) as well as genes encoding the T_reg-related cytokines transforming growth factor beta 1 (Tgfb1) (35) and interleukin 10 (Il10) (36). The T_H17 TF Rora was still expressed, whereas the T_H2 signature was weaker compared to cluster 2. Cluster 4 was largely populated by unexpanded P6⁺ cells and contained only one cell from expanded group 1 (orange). These cells showed a weak residual T_reg signature and low levels of Rora as well as Ifngr1. The T_H2 signature was largely lost. Some of these cells expressed the T_FH marker genes Slamf6 and Asap1.

Having successfully reconstructed the α and β TCRs of 32 P6⁺ CD4 T cells, we threaded the most abundant clone (Vα7D-6Jα17 Vβ13-2Jβ2-4, 10 cells) through the published crystal structure of mouse TCR YAe62 (pdb 3C60). CDR1, 2, and 3 for both α (orange) and β (green) are shown in Figure 5A. The ApoB-P6 peptide was placed in the groove of the published crystal structures of mouse I-A^b (pdb 1AIO, 2IAD, and 3C60). As expected, residues Y4, A7, S9, and S12 in the 15-mer were anchor residues (Figure 5B), with the residues S5, N6, S8, T10, and E11 pointing “up” toward the TCR heterodimer (Figure 5B). Residues S5 and N6 contacted TCRα A94 and G95 (Figure 5C). S8 made contact with TCRβ G95. T10 made contact with TCRβ R94 (Figure 5C). Thus, three contiguous amino acid residues in CDR3α (AGN) and in CDR3β (RGR) determined the specificity of this TCR for P6 in the groove of I-A^b (Figure 5D).