αβ T-cell receptors from multiple sclerosis brain lesions show MAIT cell–related features

Written consent from patient A was obtained according to the Declaration of Helsinki. The study was approved by the ethics committee of the medical faculty of the LMU Munich.

The male patient A initially presented with left-sided hemianopia in 1996. His initial cranial MRI showed a large right temporo-occipital white matter lesion, raising suspicion of malignant glioma.⁵ Two weeks after onset of his clinical symptoms, the brain lesion was neurosurgically resected. Histopathology showed an inflammatory demyelinating lesion consistent with MS. Subsequently he had a typical relapsing-remitting course of MS. He has been continuously treated with interferon-β-1a IM from the time of his third relapse in 1998 until submission of this manuscript. Figure e-1 at Neurology.org/nn gives an overview of the course of experiments. Blood samples for this study were taken in 2003, 2005, 2013, and 2014. Written consent was obtained from the donor.

The resected brain tissue was immediately frozen in liquid nitrogen and stored at −80°C until further use. Frozen sections of 10 µm were cut, mounted on positively charged slides (Superfrost Plus, Menzel, Braunschweig, Germany) for immunohistochemistry or on membrane-covered PET slides (Zeiss, Jena, Germany) for laser capture microscopy, and directly stored at −20°C or −80°C, respectively. Immunohistochemistry and microdissection was applied to tissue regions that contain lesions with high numbers of CD8⁺ T cells.

To characterize T-cell infiltrates in sections of MS brain, the following antibodies against cell surface molecules were used: mouse anti-human CD161 (1:5, 191B8, Miltenyi Biotec, Bergisch Gladbach, Germany), mouse anti-human Vα7.2 (1:5, 3C10, BioLegend, San Diego, CA), mouse anti-human CD8α (1:50, LT8, AbD Serotec, Kidlington, UK; labeled with the Cy3 MAb labeling kit, GE Healthcare, Freiburg, Germany), rabbit anti-human CCR7 (1:800, Y59, Abcam, Cambridge, UK), mouse anti-human CD45RA (1:250, HI100, BioLegend), mouse anti-human CD45RO (1:250, UCHL1, BioLegend), fluorescein isothiocyanate (FITC)-labeled mouse anti-human Vβ1 (1:100, BL37.2, Beckman Coulter, Brea, CA), rabbit anti-human CD3 (1:500, Dako, Glostrup, Denmark), and Alexa Fluor 488–conjugated mouse anti-human CD4 (1:50, RPA-T4, eBioscience, San Diego, CA). Frozen sections of CNS were thawed at room temperature, fixed in acetone, rehydrated in phosphate-buffered saline (PBS), and blocked with 2% bovine serum albumin in PBS. For staining with anti-Vα7.2 and anti-CD161 antibodies, samples were fixed in 4% paraformaldehyde for 10 minutes. Sections were then incubated for 1 hour with the primary antibodies at room temperature and subsequently—in cases where the primary antibody was not directly conjugated to a detectable fluorophore—the bound primary antibody was detected using a secondary Alexa Fluor 488–labeled goat anti-mouse IgG antibody (1:1,000, Life Technologies, Carlsbad, CA). For double staining of Vα7.2 and CD161, an Alexa Fluor 568–labeled goat anti-mouse IgG1 or an Alexa Fluor 488–labeled goat anti-mouse IgG2a antibody (1:1,000, Life Technologies) was used, respectively. If a FITC-labeled primary antibody was used, the signal was enhanced using an anti-FITC Alexa Fluor 488–conjugated secondary antibody (1:100, Life Technologies). For dual stainings with anti-CD8α, the Cy3-labeled anti-CD8α antibody was applied last. Slides were mounted using fluorescent mounting medium (Dako). Staining was visualized using an Olympus FV 1000 confocal microscope with a 40× water 1.15 NA objective or a Zeiss Axioplan 2 microscope.

Peripheral blood mononuclear cells (PBMCs) were isolated from ethylenediaminetetraacetic acid blood in 2003 by standard-density centrifugation using Percoll (Pan-Biotech, Aidenbach, Germany) and double stained with mouse anti-human Vβ1 (clone BL37.2) and CD8⁺ (clone LT8). CD8⁺Vβ1⁺ double-positive T cells were sorted using a BD FACSVantage SE flow cytometer and analyzed by single-cell PCR. Cells were either sorted directly into PCR tubes or obtained as bulk and then isolated by hand under a microscope. CD8⁺ T cells from 2003 to 2005 were analyzed by complementarity determining region (CDR) 3 spectratyping, as described previously.⁶

To isolate MAIT cells from PBMCs in 2013 and 2014, T cells were negatively isolated using the pan T-cell isolation kit (Miltenyi Biotec) and subsequently stained with the FITC-labeled mouse anti-human Vα7.2 (1:20, 3C10, BioLegend) and the allophycocyanin-labeled mouse anti-human CD161 (1:10, 191B8, Miltenyi Biotec) antibodies at 4°C for 30 minutes. Cells were then sorted using a BD FACSAriaTM2 flow cytometer. 5 × 10² CD161⁺Vα7.2⁺ cells, 6 × 10³ CD161⁻Vα7.2⁺ cells, 1 × 10⁵ CD161⁺ cells, and 2 × 10⁵ CD161⁻ cells were collected. CD4⁺ and CD8⁺ T cells were positively isolated from PBMCs in 2013 and 2014 using magnetic beads (Miltenyi Biotec).

Single Vβ1⁺CD8α⁺ T cells were isolated by microdissection as described,⁷ with the exception that we replaced the anti-Vβ5 antibody with the anti-Vβ1 antibody BL37.2. TCRs of single Vβ1⁺CD8α⁺ cells obtained by microdissection or by flow cytometry were analyzed by reverse transcription PCR as described.¹¹ PCR products were separated using a 2% agarose gel, and promising bands were excised and submitted to sequencing after isolation of DNA by the MinElute Gel Extraction Kit (Qiagen, Hilden, Germany). Throughout this manuscript, the nomenclature according to Arden et al.²⁰ is used for the TCR V region.

For the analysis of the Vα7.2 chains, RNA was isolated from sorted cells using Trizol (Life Technologies), and complementary DNA was transcribed using the Qiagen One Step RT-PCR Kit (Qiagen). Two rounds of nested PCR were performed using the primers TRAV1-2-for-out, Cα-rev-out, TRAV1-2-for-in, and Cα-rev-in.¹¹ The inner primers were extended for tags CS1 and CS2, barcodes MIDXX, and 454Ti-A and 454Ti-B adaptors according to the Access Array System 454 user guide (Fluidigm, San Francisco, CA). Samples were pooled in equimolar concentrations.

The library for sequencing all TCR β chains was prepared using 5′-RACE (Clontech, Mountain View, CA), with the gene-specific primers in the constant region CβRT¹¹ (20 µM) and the 5′ adaptor (10 µM; 5′-AAGCAGTGGTATCAACGCAGAGTACTCTT(rG)53′). This was followed by 2 nested PCRs for 20 cycles, each for 20 seconds at 94°C, 20 seconds at 60°C, and 50 seconds at 72°C using gene-specific reverse primers Cβ-mid and Cβ-in.²¹ Cβ-in was extended for barcodes and the adaptor 454Ti-B. As forward primers we used the step-out primers sa1: CTAATACGACTCACTATAGGGCAAGCAGTGGTATCAACGCAG; sa2: 454TiA_MIDXX_CTAATACGACTCACTATAGGGC. PCR products were excised from agarose gels, purified by the MinElute Kit (Qiagen), and pooled in equimolar concentrations. Quality control of the libraries and sequencing on a Roche 454 GS FLX+ platform (Roche, Basel, Switzerland) was then carried out at IMGM Laboratories GmbH, Martinsried, Germany.

We obtained a total of 183,291 TCRα reads and 182,245 TCRβ reads, which were analyzed using the IMGT/HighV QUEST platform.²² We selected for in-frame sequences that contained the complete CDR3 region with clearly distinguishable V and J genes only. Thereby we obtained 46,682 and 46,133 readable TCRα and TCRβ chains, respectively.

It was known that in patient A a particular T-cell clone expressing the Vβ1-Jβ2.3 TCRβ chain was expanded in brain lesions and persisted in peripheral blood and CSF between 1996 and 2003.^5,6 Immunohistochemistry revealed clusters of Vβ1⁺CD8⁺ cells within the CNS (figure 1A). In some of these clusters, up to 38% of all CD8⁺ T cells coexpressed the Vβ1 chain together with CD8α. This observation links the clonal T-cell expansions initially detected by CDR3 spectratyping to morphologically identifiable accumulations of T cells.

To identify the TCRα chain associated with the expanded Vβ1-Jβ2.3 chain, we double stained brain sections of the 1996 tissue for CD8α and Vβ1 and isolated single CD8α⁺Vβ1⁺ T cells by laser microdissection from brain sections (figure e-2A). In parallel, we isolated single peripheral blood CD8α⁺Vβ1⁺ T cells obtained in 2003 by flow cytometry (figure e-2B). To distinguish probable antigen-experienced Vβ1-Jβ2.3⁺ T cells from bystanders, we tested all Vβ1⁺ cells for their n(D)n-J sequences. For cells that expressed the expanded and persisting sequence Vβ1-IGRKD-Jβ2.3 (figure 1B), we analyzed the matching α chain by multiplex PCR.¹¹ We identified 15 α chains that paired with the Vβ1-Jβ2.3 chain. One T-cell clone was found 7 times in brain, and another clone was found 6 times in blood (figure 1B). All α chains share the Vα7.2 element, and even though they do not share the same Jα element, they all show homologous CDR3α regions with the consensus sequence valine, arginine (one clone: glutamine), aspartic or glutamic acid, and a medium size hydrophilic amino acid. Of note, 5 of 6 CDR loops of the 4 αβ heterodimers are identical because all clones share the identical β chain and the Vα7.2 variable element. The common motif in the CDR3 loop of the α chains therefore hints at related antigens. Strikingly, one of the α chains identified from microdissected brain-infiltrating CD8⁺ T cells represents a canonical MAIT cell α chain that contains 2 N amino acids (X) in the CDR3α loop (CAXXDSNYQLIW) and the Jα33 element.^12,15 Thus, the expanded, dominant, and persisting Vβ1 chain found in earlier studies^5,6 pairs with several distinct yet related Vα7.2⁺ α chains.

The immune cell infiltrates in lesions were dominated by CD8⁺ T cells. Four hundred twenty of 503 (84%) examined tissue-infiltrating parenchymal CD3⁺ T cells expressed CD8α (figure 2A). The parenchymal CD4:CD8 ratio was 1:8 (figure 2B). Expression of the naive T-cell marker CD45RA was found on only 11% of tissue-infiltrating CD8α⁺ cells (figure 2C), whereas 84% expressed the memory marker CD45RO (figure 2D). Of all parenchymal CD8α⁺ cells, 60% coexpressed the homing receptor CCR7. Double staining of CCR7 together with CD45RA or CD45RO revealed the following phenotypes in decreasing frequency: central memory (CD45RO⁺CCR7⁺), effector memory (CD45RO⁺CCR7⁻), and naive T cells (CD45RA⁺CCR7⁺), which accounted for only 6% of all CCR7⁺ cells (figure 2E). Sixty-five percent of CCR7⁺ cells coexpress CD45RO, and 62% of all CD45RO⁺ cells express CCR7 (figure 2F). Thus, the brain-infiltrating T cells predominantly show a CD8α⁺CD45RO⁺CCR7⁺ central memory or CD8α⁺CD45RO⁺CCR7⁻ effector memory phenotype.

To investigate phenotypic features of the MAIT-related cells, we stained brain sections for the expanded Vβ1⁺ TCRβ chain and MAIT cell–associated markers Vα7.2, CD161, and CD8α. Forty-two percent of the parenchymal Vα7.2⁺ cells coexpressed Vβ1; however, only 13% of Vβ1⁺ T cells expressed Vα7.2 (figure 3A). Seventy percent of the Vα7.2⁺ cells also expressed CD8α (figure 3B), and 69% of Vα7.2⁺ cells coexpressed CD161 (figure 3C), identifying them as MAIT cells.¹³ Of all CD161⁺Vα7.2⁺ MAIT cells, 74% were found in the brain parenchyma. Comparing the number of parenchymal CD161⁺Vα7.2⁺ cells to the total number of parenchymal CD8α⁺ cells in consecutive sections, we found that only about 1% of all CD8α⁺ cells belong to the CD161⁺Vα7.2⁺ MAIT cell subset. However, CD161 was found on 21% of CD8α⁺ T cells. Conversely, about 80% of the CD161-expressing cells were CD8α⁺ (figure 3D). Furthermore, about 80% of Vβ1⁺ T cells expressed CD161⁺ (data not shown). This underlines the fact that many of the CD161⁺ cells are not MAIT cells.²³

To follow the longitudinal development of the peripheral T-cell repertoire from 1996 to 2014 (figure e-2), we analyzed the 1996 brain sample as well as different subsets of peripheral T cells from 2013 to 2014 by pyrosequencing.

Tracking of TCR chains identified by single-cell PCR in the 1996 biopsy specimen revealed that the Vβ1-Jβ2.3 chain that was expanded in brain and blood samples from 1996, 2001, and 2003^5,6 was still detectable by pyrosequencing in peripheral T cells 18 years after clinical onset, but it was no longer expanded. Furthermore, the canonical Vα7.2-CAVRDSNYQLIW-Jα33 MAIT cell clone found in the 1996 brain sample by single-cell PCR was still present in the 2014 CD161⁺Vα7.2⁺ MAIT PBMC subset, as found by pyrosequencing.

Next, we compared the CDR3 repertoires of Vα7.2⁺ chains. We show the distribution of the N and J regions of the 1996 brain specimen in figure 4A. One of the 3 dominant clones is a canonical MAIT cell Vα7.2-Jα33 clone, although it differs from the clone found in the single cells. In peripheral blood from 2014, clonal expansions were found in only the CD161⁺Vα7.2⁺ subset, with the MAIT cell canonical and noncanonical clones dominating (blue in figure 4B). In contrast, the CD161⁻Vα7.2⁺ subset is polyclonal (figure 4C). We also detected clonal expansions including some typical sequences for MAIT cells in the entire pool of CD8⁺ T cells (figure 4D), whereas CD4⁺ T cells were polyclonal (figure 4E).

We then set out to identify distinct clones shared between these different subsets (figure 4F). Within the TCR Vα7.2 repertoire, several clones that were present in the 1996 brain sample can still be found in blood T cells obtained 18 years after onset, especially in the CD161⁺Vα7.2⁺ MAIT cell subset. The greatest overlap was observed between the CNS sample and the PBMC MAIT cell sample from 2014, in which 2.4% of all sequences were identical at the amino acid level. It is notable that the canonical Vα7.2-CAVRDSNYQLIW-Jα33 sequence could also be detected in T-cell subsets other than the MAIT cells.