Cross-reactive T-Cell Receptors in Tumor and Paraneoplastic Target Tissue

Background  According to established criteria, paraneoplastic encephalomyelitis with adrenal neuroblastoma comprises a definite paraneoplastic neurologic syndrome.

Objective  To detect T-cell clones that cross-react against antigens shared between tumor and nervous system.

Design  Case study.

Setting  Academic research.

Patient  A 22-year-old woman having paraneoplastic encephalomyelitis with adrenal neuroblastoma.

Main Outcome Measures  We compared the T-cell receptor repertoires expressed in blood, cerebrospinal fluid, and neuroblastoma tumor tissue using complementary determining region 3 (CDR3) spectratyping and clone-specific polymerase chain reaction.

Results  The T-cell receptor repertoire in cerebrospinal fluid was narrow compared with that in tumor and blood. Four T-cell clones from different tissues had identical T-cell receptor β chains. Remarkably, the chains showed identical amino acid sequences but different nucleotide sequences.

Conclusions  These T cells represent ontogenetically distinct clones but share functionally identical receptors. They recognize the same antigen in nervous system and tumor tissue and represent an attractive target for selective therapy.

The exact mechanisms of neuronal damage in paraneoplastic neurologic syndromes (PNS) are unknown.^1-4 Many patients mount a high-titer immunoglobulin response to specific onconeuronal antigens^5-7 that are also ectopically expressed in the associated tumors. Such antibody responses have been well characterized. However, it is difficult to identify cross-reactive T-cell responses,^5,6 although the transfer of T cells specific for an onconeuronal antigen can induce encephalitis in rats.⁸ Presumably, the reasons are that PNS are rare, tissue specimens are scarce, and few T cells are detectable in situ.

We studied a case of paraneoplastic encephalitis with adrenal neuroblastoma in which we compared the T-cell receptor (TCR) repertoires in different tissue compartments. We identified clonal T-cell expansions using complementary determining region 3 (CDR3) spectratyping⁹ and tracked individual TCR β chains using clone-specific polymerase chain reaction (PCR). We detected identical TCR chains in the adrenal tumor, cerebrospinal fluid (CSF), and CD8⁺ compartment of peripheral blood from the patient, indicating that these T cells were directed against the same target antigen. T cells recognizing peptides from PNS antigens have been detected in the peripheral blood of patients with PNS.^10-12 To our knowledge, this is the first description of TCR cross-reactivity between the (protective) antitumor response and the concomitant adverse paraneoplastic response against nervous system tissue that was detected in corresponding tissue samples.

We studied tissue specimens from a 22-year-old woman having paraneoplastic encephalomyelitis with adrenal neuroblastoma, which is a definite PNS according to established criteria.¹³ Nine months after the first clinical signs (blurred vision), she developed progressive gait disturbance and seizures. Cranial magnetic resonance imaging revealed marked cerebellar atrophy. The CSF analysis showed oligoclonal banding. Anti-Hu antibodies were detected in serum and CSF at dilutions of 1:1920 and 1:30, respectively. Tumor search revealed a para-aortal neuroblastoma. Tumor resection was followed by chemotherapy and stem cell transplantation. Tissue samples were obtained with the patient's informed consent and were stored at −80°C. The study was approved by institutional review boards of the Ludwig-Maximilians University, Munich, Germany, and was conducted according to the tenets of the Declaration of Helsinki. Subsequently, the patient had tumor recurrence requiring reoperation and chemotherapy. Eight years after tumor therapy, the patient is neurologically stable without evidence of tumor recurrence.

CDR3 spectratyping of TCR β chains has been previously described.^9,14 Fragment lengths were analyzed on a DNA sequencer (ABI377; Applied Biosystems, Darmstadt, Germany). Polyclonal populations show gaussian distributions. Clonal expansions appear as distinct peaks. We considered only signals that exceeded the baseline scatter by a factor of more than 5. Notable Vβ-Jβ combinations that were expanded in at least 2 different tissues were reamplified and sequenced directly. Candidate clones were regarded as monoclonally expanded only if readable sequences were obtained.

Clone-specific PCRs were performed using the PCR products of the V-C spectratyping as templates. We used the forward primer Vβ8-forward¹⁴ and the reverse primer Jβ1.5-rev¹⁴ or Jβ1.5-N3–rev as follows: 5′-GCTGGGGCTGATTGCTA-3′ (NDN nucleotides are underlined). After a 5-minute incubation at 94°C, 35 PCR cycles were run at 58°C, 94°C, and 72°C for 1 minute each. The PCR products were cloned (TOPO-TA; Invitrogen, Karlsruhe, Germany), and individual bacterial clones were analyzed.

Table 1 lists the numbers of single peaks and polyclonal distributions detected in tumor, CSF, and peripheral blood. The quotients of the numbers of single peaks and polyclonal distributions in each tissue are semiquantitative measures for the widths of the repertoires. The quotients are approximately 0.2 in all tissues, indicating predominantly polyclonal TCR repertoires. As an exception, we observed a quotient of 10.7 for T cells in CSF, indicating few, but strongly, expanded clones.

Table 2 lists how often single peaks at identical Vβ-Jβ combinations were observed in 2 tissues or more. We detected 18 T-cell clones with identical Vβ-Jβ combinations in several combinations of different tissues. With one important exception (described herein), sequencing revealed that none of the clones were identical, as the NDN nucleotide sequences and the deduced amino acid sequences of the NDN regions were all different.

The left panels of Figure 1 show spectratyping peaks for the combination Vβ8-Jβ1.5. The right panels show the electropherograms of the sequencing reactions. We found a monoclonal expansion in CSF (Figure 1A) as indicated by a single peak and a readable sequence. In tumor tissue, we found a skewed polyclonal population that had a prominent peak at the relevant CDR3 length (Figure 1B). In the CD8⁺ compartment of peripheral blood, we again detected a single peak (Figure 1C). Sequencing revealed a dominant sequence with the first amino acid of the N region (threonine) coded by the nucleotides ACC, as well as a subdominant sequence with A(G)C coding for serine. This yields precisely the same NDN-region amino acid sequence as that detected in CSF (Figure 1A). However, the CSF clone and the subdominant clone in blood are identical only at the amino acid level, as proline is coded by the nucleotide CCC in CSF but by CCT in blood (Figure 1A and C, green arrows).

To detect these clones in tumor tissue, we used clone-specific PCR, which may identify relevant clones even in the presence of other clones with different CDR3 sequences. We used the following 2 reverse primers: primer Jβ1.5-rev hybridizes completely within the J region, whereas the clone-specific primer Jβ1.5-N3–rev includes the last 3 nucleotides of the NDN region but does not hybridize to the nucleotides under investigation (Figure 2, left panels). Figure 2A (right panels) shows that we identified 2 of 24 clones with the Jβ1.5-rev primer, which yielded the identical amino acid sequence as that identified by CDR3 spectratyping in the CSF (Figure 1A). Amplification with the clone-specific primer Jβ1.5-N3–rev yielded 35 of 41 clones with the correct amino acid sequence (Figure 2B). All these clones had different silent exchanges compared with the CSF sequence. In total, we identified 4 different silent nucleotide exchanges with serine and proline coded by AGC-CCC, AGC-CCT, AGT-CCT, and AGT-CCG. Of note, we could not detect the CSF resident clone (AGC-CCC [Figure 1A]) in blood or tumor tissue. However, the subdominant clone found in CD8⁺ blood cells (A[G]C-CCT [Figure 1C]) was also found in tumor tissue.

In an informative patient having paraneoplastic encephalitis with adrenal neuroblastoma, we observed several expanded T-cell clones in all tissues on a polyclonal background. As expected, the overall TCR repertoires in peripheral blood and tumor were predominantly polyclonal (Table 1 and Figure 1B), whereas the T-cell repertoire in the CSF was significantly narrower. This would fit with the assumption that the paraneoplastic response against CNS tissue is initiated by a few cross-reactive clones.⁵ Indeed, we identified such cross-reactive T cells in tumor, CSF, and blood from this patient. Intriguingly, 4 different T-cell clones had identical TCR β-chain amino acid sequences but different nucleotide sequences. Because the hypervariable CDR3 regions are highly characteristic for each T-cell clone, the probability that 2 TCR chains or more are identical “just by chance” is close to zero. One clone was found exclusively in the CSF. Three “sister clones” were detectable in the adrenal tumor; 1 of these clones was also present in the CD8⁺ compartment of peripheral blood. These observations strongly suggest that these T-cell clones recognized an antigen shared between the adrenal neuroblastoma and the paraneoplastically affected CNS tissue. We do not know whether these clones recognize Hu or another target antigen, but our results strongly suggest a convergence of immune recognition in tumor and CSF. Because the clone from peripheral blood was CD8⁺, it is likely that the T-cell clones from tumor and CSF also belong to the CD8⁺ subset, indicating that the pathogenic paraneoplastic response was at least partially mediated by cytotoxic T cells. Nevertheless, further characterization of the phenotype and biologic activity of such clones will be required to prove their direct relevance to PNS.

Identification of such cross-reactive T clones may represent a new therapeutic perspective. If it was possible to selectively eliminate these cells, the paraneoplastic autoimmune response might be diminished or abolished, whereas most tumor-specific cells would remain largely unaffected.

Correspondence: Klaus Dornmair, PhD, Department of Neuroimmunology, Max Planck Institute for Neurobiology, Am Klopferspitz 18, D-82152 Martinsried, Germany (dornmair@neuro.mpg.de).

Accepted for Publication: October 1, 2008.

Author Contributions:Study concept and design: Pellkofer, Voltz, Goebels, Hohlfeld, and Dornmair. Acquisition of data: Pellkofer and Voltz. Analysis and interpretation of data: Pellkofer, Voltz, Hohlfeld, and Dornmair. Drafting of the manuscript: Pellkofer, Hohlfeld, and Dornmair. Critical revision of the manuscript for important intellectual content: Pellkofer, Voltz, Goebels, Hohlfeld, and Dornmair. Obtained funding: Voltz and Dornmair. Administrative, technical, and material support: Pellkofer, Voltz, Goebels, Hohlfeld, and Dornmair. Study supervision: Voltz, Hohlfeld, and Dornmair.

Financial Disclosure: None reported.

Funding/Support: This study was supported in part by the Hermann and Lilly Schilling Foundation and by grants SFB 571-A1 and SFB 571-D7 from the Deutsche Forschungsgemeinschaft.

Additional Contributions: Ingrid Eiglmeier and Joachim Malotka provided expert technical assistance. Edgar Meinl, MD, Tania Kümpfel, MD, and Hinrich Abken, MD, gave helpful comments on the manuscript.