Comparison of T-cell receptor repertoire restriction in blood and tumor tissue of colorectal cancer patients

Several immunotherapeutic approaches rely on antigen-specific T-cells. Restrictions in the T-cell receptor (TCR) repertoire were reported as indicator of anti-tumor cytotoxic T-lymphocyte (CTL) response in various tumor entities. It is unclear yet whether a TCR restriction in peripheral blood mirrors the tumor compartment. We compared the expression of TCR Vβ-families for the quantification of TCR repertoire alterations in blood and tissue samples from patients with colorectal carcinoma. Blood samples from patients with colorectal carcinoma and healthy volunteers and tissue samples of normal colonic mucosa and colorectal carcinoma were analyzed. Relative Vβ-family quantification was performed based on quantitative reverse transcribed PCR. Standard deviation and average mean of the single families were determined. Two variables describing the degree of Vβ-repertoire restriction were defined. Forty-eight blood samples and 37 tissue samples were analyzed. TCR repertoire restriction was higher in blood of tumor patients than in blood of healthy controls (p < 0.05). No difference in the degree of TCR repertoire restriction was found between carcinoma and unaffected colon tissue. We found no corresponding elevated TCR families among the different compartments blood, normal colon, and carcinoma tissue of the same patient. In conclusion, we observed a repertoire restriction in peripheral blood as well as in tumor tissue of cancer patients. However, in tumor tissue, repertoire alterations were comparable to normal mucosa, suggesting compartment-specific TCR distribution rather than alterations due to tumor-T-cell interaction questioning the presence of highly restricted clonal T-cell expansions in colorectal cancer as they have been described in other, assumingly more immunogenic tumor entities.


Background
Understanding the interaction between tumor and immune system might help improving immunotherapeutic approaches for malignant diseases. T-cells directed against tumor associated antigens (TAA) could play a key role in the surveillance of and in the defense against tumor cells [1]. In fact, spontaneous T-cell responses against TAAs have been described in peripheral blood, lymph nodes, and bone marrow of patients with various malignant diseases prior to immunotherapy [2].
In colorectal cancer (CRC), spontaneous T-cell responses against several TAAs have been detected in peripheral blood, particularly in patients with metastatic disease [3,4]. No evidence was found that these spontaneous, peripheral TAA-specific T-cells have an impact on survival of CRC patients [5]. Therefore, the focus of interest has moved to tumor-infiltrating T cells. CD8+ T-cell infiltration of CRC is known to be associated with a better prognosis, but it is still unknown whether these infiltrating T cells, in fact, represent expanded tumor specific T-cell clones [6][7][8][9][10][11][12][13].
In case of unknown or multiple epitopes, the analysis of TCR repertoire both by FACS and PCR based methods offers the opportunity to detect oligoclonal expansion of specific T-cells [14][15][16]. The dimeric transmembrane Tcell receptor (TCR) is the central mediator of epitope specific cytotoxic T-cell activation. Consisting of an α-and a β-chain in most of the cases, diversity is generated during T-cell evolution by recombinations of the gene segments V (variable), in case of the β-chain D (diversity), and J (joining) to a constant chain gene C [17]. V-genes are grouped in families consisting of genes with sequence homology of at least 50% [18]. For analysis of the TCR repertoire, the β-chain is often preferred because of the lower number of families even if a higher overall variability of sequence compared to the α-chain has been described [19]. Alterations in TCR repertoire can be evaluated either by length or sequence analysis of the highly variable part of the α-or β-chain for each V-family [14,[20][21][22] or by quantification of the single families by southern blot, FACS, or quantitative reverse transcribed PCR (qRT PCR) [23][24][25][26][27].
In cancer research, a restricted TCR repertoire has been found at the tumor site of various malignant diseases [28][29][30][31][32][33][34][35][36], and in case of melanoma, a highly restricted repertoire may be linked to regression during cytokine therapy [37]. However, it is still a matter of debate whether a restricted TCR repertoire in peripheral blood of tumor patients exists and whether such a peripheral restriction mirrows oligoclonal expansions of specific T-cells in the tumor compartment [36,[38][39][40][41][42].
We used a qRT PCR-based relative Vβ-family quantification approach [27] for analysis of TCR Vβ-family expression. Especially in the gut, lymphocytes bearing γδ TCR are abundant, which are potentially involved in an antitumoral response in an MHC-independent manner [43]. Assessing Vβ-family restriction, clonal expansions of γδ T-cells are not addressed. Aim of the study was the application of mathematical markers to describe the global restriction of the αβ TCR repertoire in the different compartments rather than the detection of single expanded T-cell clones. From this general point of view we evaluated whether or not significant differences of TCR repertoire restriction can be detected in samples from carcinoma patients and healthy controls as well as in tumor tissue compared to unaffected colonic mucosa.

Specimen collection
Peripheral blood samples were drawn from patients and healthy volunteers. Tissue samples both of carcinoma and unaffected mucosal tissue were collected from patients affected by CRC undergoing tumor resection. RNA was extracted from the macroscopic center of the tumor and from unaffected colonic mucosa at least 5 cm from the macroscopic border of the malignant lesion. Age, sex, and in CRC patients TNM and UICC stages were assessed. Both patients and controls had given informed consent for the use of their specimens before sampling.

Quantification of TCR expression
For determination of general TCR expression, the Cαchain was quantified on a LightCycler ® instrument (Roche, Basel, Switzerland). Values were normalized using the low-abundance housekeeping gene porphobilinogen deaminidase (PBGD) as previously described [27]. For each cDNA synthesis reaction (+RT), a control reaction without reverse transcriptase was performed (-RT control). Samples with -RT/+RT ratio >0.1 for Cαchain or PBGD indicating DNA contamination and/or RNA degradation were excluded from further analysis. The aim of our study was the detection of TCR repertoire restriction due to T-cell expansions associated with significant TCR expression of the expanded clones. Low overall TCR transcription in tissue could lead to putative oligoclonality due to the generally high sensitivity of PCR based approaches. Consequently, samples with HAC/ PBGD < 0.1 were excluded from analysis to avoid this bias.

Relative quantification of Vβ-families
Relative quantification of the expression of a single TCR Vβ-chain was performed as described [27]. Shortly, qRT PCRs were performed with a universal reverse primer and TaqMan probe, both annealing at the constant part of the β-chain (Cβ), and 28 Vβ-family-specific forward primers (Vβ-family 1 to 24, Vβ-families 5, 6, 12, 13 subdivided into two subgroups each). Slope of each family specific reaction was estimated analyzing a dilution series spanning three orders of magnitude of a cDNA mixture of diagnostic samples. Calculation of the relative concentration P j [%] of a Vβ-family j was carried out using the formula with Cp [amplification cycles] is the Crossing point and

Normalization, quantification of Vβ-restriction
The approach calculating relative concentrations of different Vβ-families regarding slopes and crossing points, as a matter of fact, leads to per-sample normalized values. To accommodate the fact, that in tumor immunological settings, a mono-or oligoclonal T-cell response is postulated [16], which is supposed to be associated with the elevation of only a single or very few increased families, we defined a second marker n(F). For each sample k, the number n(F) of families was determined, which were expressed higher than the mean percentage plus two standard deviations SD j (P' j > 2). In contrast to CD depicting the sum of expression deviations from the average of all Vβ-families, only significantly elevated families were relevant for the value of the second marker n(F). The applied normalization procedure is depicted exemplary in Figure 1.

Statistical methods
All statistical tests were performed as two-tailed tests and a p-value < 0.05 was considered significant. Correlations were tested by calculation of the Pearson coefficient.
Mann-Whitney U test was applied to unpaired samples, Wilcoxon test was used for paired samples.

Patients and specimens
Fifty-one samples from peripheral blood ( Table 1.
To test the reliability of our normalization approach, we calculated the theoretically expected percentage of samples with at least one family elevated more than two SD

TCR Cα expression in blood and tissue specimens
As expected, Cα (HAC/PBGD) was expressed significantly higher in peripheral blood than in tissue (p < 0.0001). HAC/PBGD was higher in blood samples of healthy volunteers compared to carcinoma patients; but no difference in Cα-expression was found between unaffected colon and tumor tissue in a paired analysis ( Figure  2). These results have to be interpreted with caution because of the fact that HAC/PBGD was used as criteria to select samples for further repertoire analysis. Without excluding samples with HAC/PBGD < 0.1, Cα concentration was lower in the carcinoma samples than in samples of unaffected mucosa (paired, p = 0.047).  According to the differences of expression between blood and tissue specimens, analyzing all evaluable samples irrespective of origin, we found a negative correlation between HAC/PBGD and CD (Figure 3) as well as HAC/PBGD and n(F) (not shown). However, separate analysis of the samples originating from blood and originating from tissue samples resulted in no significant correlations (Figure 3 and not shown). In the statistical analysis, both CD and n(F) were higher in PBMCs from cancer patients than in PBMCs from healthy volunteers (unpaired, p CD = 0.0316, p n(F) = 0.0195, Figure 2B).

Vβ repertoire in tissue samples of CRC patients
Twenty-four samples of normal colon tissue and 18 carcinoma samples were evaluable. The statistical analysis of both CD and n(F) comparing all tissue samples (carci- noma and unaffected colon) with all blood samples (healthy controls and carcinoma patients) showed a significantly higher repertoire restriction in tissue than in blood (unpaired, p CD < 0.001, p n(F) < 0.001, Figure 4A). However, no difference was observed between samples of unaffected colon and carcinoma tissue (paired, p CD = 0.471, p n(F) = 0.2783, Figure 4C).

Comparison of elevated families in different compartments
For 16 patients, samples of tumor and tumor free colon were available. In the carcinoma samples, altogether 30 families were elevated (0-5 families per patient). In the corresponding samples of unaffected colon tissue, 14 families were elevated (0-4 families per patient). In only two of these patients a single family was elevated in both tumor tissue and corresponding unaffected colon. For nine CRC patients, blood and tumor samples were available. Four of these patients had elevated families in peripheral blood, but none of these families was elevated in the corresponding tumor sample. Please refer to figure 4.

Discussion
In this study, TCR Vβ-family repertoire restrictions in blood and tissue of patients with colorectal carcinoma were compared using high throughput relative quantifi-cation of TCR Vβ-families based on qRT PCR technology. While a multitude of mathematical approaches have been established to describe the degree of repertoire restriction in single V-families from spectratype/ immunoscope data (reviewed in [45]), to our knowledge, a model to describe global TCR repertoire restriction based on V-family quantification has not been published so far. Colorectal carcinoma is thought to be of limited immunogenicity. Nevertheless, global Vβ-repertoire restriction degree in blood reflected by the CD and n(F) values was significantly higher in carcinoma patients confirming the results of former studies investigating Vβ-repertoire restrictions in blood of patients with other tumor entities, which have been interpreted as indication for the induction of specific clones reactive to autologous tumor [26,42]. In fact, this TCR repertoire restriction in tumor patients might be attributed to cumulative Vβ-alterations caused by discrete epitope-specific T-cell expansions triggered by several different antigens. Because of the epitope-independency of the assay, unspecific antigen confrontation due to barrier disruption as postulated in the context of other tumor entities [28,32] causing additional repertoire alterations can not be excluded. We observed a difference of Cα-chain expression between healthy controls and carcinoma patients, which could have impact on the TCR Vβ restriction grade. It remains elusive whether or not this difference is a result of the malignant disease or the higher median age of the patient cohort compared to the healthy control group.
Corresponding significant TCR Vβ-family elevations in blood and tumor tissue were not found in the present study. From previous analyses in colorectal cancer, we know, that TAA-specific T-cells circulated with frequencies lower than 1% of CD8+ cells without proven clonality [3]. As anticipated, the proportion of peripheral TAAspecific T-cells was too low to be detected by TCR Vβquantification due to physiological variation of the relative Vβ-family expressions. Assuming several TAAs, expansions of T-cell clones would affect various Vβ-families additionally reducing sensitivity of the relative Vβquantification regarding single family percentages. It is purely speculative to assume that such a specific T cell clone might be trapped at the tumor site and is therefore, not detectable in the peripheral blood.
In carcinoma tissues, we did not observe an elevated repertoire restriction compared to corresponding tissue of unaffected colon. Infiltration of CTLs in colon carcinoma tissue had been identified as prognostic factor suggesting TAA associated CTL activation [12]. Interestingly, we found no difference in the expression of Cα comparing healthy colon and carcinoma tissue. These inconsistent observations could be explained by recent results of Salama et al., who showed, that CD8+ CTL Figure 3 Correlation between relative Cα expression (HAC/PBGD) and cumulative deviation CD. According to the higher HAC/PBGD expression in blood samples compared to tissue specimens, we found a strong negative correlation between HAC/PBGD and CD analyzing all specimens irrespective of origin (regression line, R p = -0.4004, p = 0.0003). Regarding samples from blood (empty diamonds) and tissue (filled diamonds) separately, no correlations between the mentioned parameters were determined (R p, blood = -0.1108, p blood = 0.4585, R p, tissue = 0.01344, p tissue = 0.9286). density is reduced but CD4+/CD25+/FoxP3+ T regulator cells are elevated in colon cancer compared to normal colon tissue [46,47]. The fact that clonal CTL expansions are not principally associated with restrictions in Vβfamily usage of the T regulatory cell population of the same compartment [42,47] might explain the absence of TCR restriction differences between healthy colon and CRC tissue. Independently of these considerations, we have to conclude that the present molecular quantitative TCR analysis is not suitable for the identification of local expansions in colorectal cancer tissue (if they exist) compared to normal colon tissue. This may differ in other compartments and/or for other tumor entities.
Regarding p-values comparing the different sample groups (blood of healthy controls, blood of carcinoma patients, unaffected colonic mucosa, carcinoma tissue) one has to keep in mind the differences in sample sizes and applied tests. The highly significant difference in CD between blood and tissue samples but the absence of a significant difference between unaffected colon and tumor tissue does not necessarily mean hat there would be no difference to detect if we would compare the same sample size of tissue samples as we used for comparing tissue with blood samples (48 vs 42 instead of 24 vs 18). Regarding the repertoire restriction in blood (healthy vs carcinoma patients) and tissue (unaffected colon mucosa vs carcinoma tissue), the results of the two statistical tests cannot directly be compared because of differences in distribution, range, and dependency of the data sets. However comparing tumor tissue to unaffected mucosa, we observed an even lower median value in carcinoma tissue. Consequently, the absence of a significantly higher Vβ restriction in tumor tissue must not be attributed to the slightly smaller sample size compared to the blood samples.
Taken together, we observed an increased TCR repertoire restriction in blood of colorectal carcinoma patients compared to blood of healthy controls using qRT-PCR Figure 4 Statistical comparison of repertoire restriction degree measured in CD and n(F). Compared were samples from blood with tissue (both normal colon and carcinoma, A), blood from healthy controls with blood from carcinoma patients (B), and tumor free colon with carcinoma samples both originating from CRC patients (C). * p < 0.05, ** p < 0.01.
based Vβ-family quantification. A similar degree of TCR restriction in colorectal carcinoma tissue in comparison to normal colon tissue contrasts with the phenomenon of high proliferative oligoclonal expansions of specific T-cell populations as described for highly immunogenic malignancies such as melanoma [16].