Quantifying Distribution of Flow Cytometric TCR-Vβ Usage with Economic Statistics

Measuring changes of the T cell receptor (TCR) repertoire is important to many fields of medicine. Flow cytometry is a popular technique to study the TCR repertoire, as it quickly provides insight into the TCR-Vβ usage among well-defined populations of T cells. However, the interpretation of the flow cytometric data remains difficult, and subtle TCR repertoire changes may go undetected. Here, we introduce a novel means for analyzing the flow cytometric data on TCR-Vβ usage. By applying economic statistics, we calculated the Gini-TCR skewing index from the flow cytometric TCR-Vβ analysis. The Gini-TCR skewing index, which is a direct measure of TCR-Vβ distribution among T cells, allowed us to track subtle changes of the TCR repertoire among distinct populations of T cells. Application of the Gini-TCR skewing index to the flow cytometric TCR-Vβ analysis will greatly help to gain better understanding of the TCR repertoire in health and disease.


Introduction
A broad T cell receptor (TCR) repertoire is considered critical for optimal immunity against a wide variety of antigens [1,2]. Contraction of the TCR repertoire is associated with poor vaccine responses in aged individuals [3], progression to AIDS in HIV-infected patients [4], and poor survival in cancer patients [5]. Assessment of the T cell receptor (TCR) repertoire is therefore relevant to researchers in many fields of medicine.
Several techniques are currently used to study the TCR repertoire [6]. Moderate to high resolution assessment of the TCR repertoire is provided by PCR-based methods, such as TCR spectratyping and sequencing. As these methods are relatively labor-intensive and preferably require cell-sorting of highly pure T cell populations, many researchers turn to flow cytometry [6]. Flow cytometry quickly measures the proportional TCR-Vβ usage in multiple T cell subsets on a per-cell basis, without the need for cell-sorting. Although this assessment can provide helpful information, an accurate and reliable way for analyzing the flow cytometric data on the TCR repertoire is currently lacking.
Here, we introduce economic statistics to improve the analysis of flow cytometric data on TCR-Vβ usage. We noticed that the distribution of TCR-Vβ families among T cells resembles the distribution of income among people (Fig 1A-1C). Economists typically study the distribution of income by constructing Lorenz curves and calculating the Gini index. The Gini index, with scores ranging from 0 to 100, is a direct measure of income distribution [7][8][9]. By applying the Gini index to the flow cytometric TCR-Vβ analysis, we could directly measure the distribution of 24 TCR-Vβ families among multiple, well-defined T cell subsets. In this context, low Gini index values indicated equal distribution of TCR-Vβ families (i.e. broad repertoire), whereas high values pointed to unequal distribution of TCR-Vβ families (i.e. repertoire skewing). Although the Gini index has been used in TCR sequencing studies [10,11], we here demonstrate that the Gini index, hence referred to as the Gini-TCR skewing index, also aids the analysis of flow cytometric data on TCR-Vβ usage. Importantly, the Gini-TCR skewing index allowed us to detect subtle changes of the TCR repertoire among multiple, well-defined T cell subpopulations.

Methods Subjects
Heparinized blood samples were acquired from 27 healthy volunteers. Eight men and 19 women were included (age range 22-81). Samples of 5 children (age 9 years) undergoing DTa-P-IPV vaccination were also collected. Written informed consent was obtained from adult

Calculation of Gini-TCR skewing index
Gini-TCR skewing index values were calculated for T cell subsets of individual blood donors with use of the Gini index, which is generally used to measure income distribution [7,8]. A Microsoft Excel file allowing automatic calculation of Gini-TCR skewing index from percentages of 24 TCR-Vβ families is provided in the Supporting Information (S1 File). A simplified overview to calculate the Gini-TCR skewing index is provided (S2 Fig). Briefly, proportions of all 24 TCR-Vβ families within a T cell subset were arranged from small to large. To construct Lorentz curves, the cumulative proportions of all TCR-Vβ family proportions were normalized to a total of 100%. Lorenz curves for individual blood donors were plotted with the cumulative percentage of the 24 TCR-Vβ families studied on the x-axis, and the cumulative proportion of CD4 or CD8 T cells that were covered by these 24 TCR-Vβ families on the y-axis. Next, the surface under the Lorenz curve was subtracted from the surface under the equidistribution line (i.e. line of perfectly equal distribution). To calculate the Gini-TCR skewing index, the surface between the equidistribution line and the Lorenz curve was divided by the total surface under the equidistribution line (100 Ã 100 / 0.5 = 5000) and then multiplied by 100. An example of the complete calculation of a Gini-TCR skewing index is shown in Table 2.

Hierarchical clustering analysis
Gini-TCR skewing index values of all T cell subsets from individual donors were plotted in a heat map using Microsoft Excel 2007. Colors were appointed based on absolute Gini-TCR skewing index values. Subsequently, hierarchical clustering analysis of T cell subsets and blood donors was performed in Genesis 1.7.6 (Genomics and Bioinformatics Graz) using Euclidean distance and compete linkage clustering.

CMV ELISA
CMV-specific IgG was determined in serum samples using an in-house ELISA. 96-well ELISA plates (Greiner) were coated with lysates of CMV-infected fibroblasts overnight. Lysates of non-infected fibroblasts were used as negative controls. Following the coating, dilutions of serum samples were incubated for 1 hour. Goat anti-human IgG was added and incubated for 1 hour. Samples were incubated with phosphatase for 15 minutes and the reaction was stopped with NaOH. The plates were scanned on a Versamax reader (Molecular Devices). A pool of sera from 3 CMV-seropositive individuals with known concentrations of CMV-specific IgG was used to quantify levels of CMV-specific IgG in the tested samples.

Vaccination
Nine year old children in good health were vaccinated with a combination vaccine against diphtheria, tetanus, acellular pertussis and inactivated poliovirus (DTaP-IPV). Peripheral blood mononuclear cells were isolated by density centrifugation before vaccination, as well as 28 days and 1 year after vaccination. Samples were viably stored at -135°C with RPMI + 10% FCS and 10% DMSO. To perform flow cytometric staining, the samples were thawed in a 37°C water bath and washed twice with RPMI + 10% FCS.

Statistics
Gini-TCR skewing index values between two T cell subsets within individuals were compared with the Wilcoxon signed rank test. In case multiple T cell subsets were compared, the Wilcoxon signed rank test was preceded by the Friedman test. Gini-TCR skewing index values in CMV seropositive and seronegative individuals were compared with the Mann Whitney U test. Two-sided p values < 0.05 were considered statistically significant. Statistical analysis was performed with IBM SPSS Statistics 20 (SPSS, Chicago, IL, USA).

Calculation of the Gini-TCR skewing index
To demonstrate that flow cytometry and economic statistics can be combined, we started by calculating Gini-TCR skewing index values for CD4 and CD8 T cells from healthy individuals. Therefore, we first measured the proportional usage of 24 TCR-Vβ families in CD4 and CD8 T cells with a commercially available panel of monoclonal antibodies (S1 Fig). In accordance with previous reports, the antibody panel recognized TCR-Vβ families on 66 ± 1% (mean ± SEM) of CD4 T cells and 60 ± 2% of CD8 T cells [12,13]. At this stage, it was difficult to conclude whether the distribution of TCR-Vβ families differed between CD4 and CD8 T cells (Fig 2A). Next, the proportional TCR-Vβ usage of CD4 and CD8 T cells from individual blood donors was plotted in Lorenz curves with the cumulative percentage of the 24 TCR-Vβ families studied on the x-axis, and the cumulative proportion of CD4 or CD8 T cells that was covered by these 24 TCR-Vβ families on the y-axis (Fig 2B and S2 Fig). A Lorenz curve far from the equidistribution line refers to unequal distribution and skewing of the TCR repertoire. Finally, Gini-TCR skewing index values were calculated for CD4 and CD8 T cells of individual blood donors (S2 Fig). In some individuals, Gini-TCR skewing index values were markedly higher for CD8 T cells than CD4 T cells, whereas the values were similar in others (Fig 2C). This difference could be explained by cytomegalovirus (CMV) infection, as only CD8 T cells of CMV-seropositive individuals demonstrated increased Gini-TCR skewing index values ( Fig 2D). Furthermore, we noted that variation of Gini-TCR skewing index values in CMVseronegative could be explained by age, as we observed a substantial, aging-associated increase in Gini-TCR skewing index values among CD8 T cells of CMV-seronegative individuals (Fig 2E). In contrast, the TCR-Vβ repertoire of CD8 T cells was already skewed in young CMV-seropositive donors and did not modulate further with age. In addition, Gini-TCR skewing index values only modestly increased with age among CD4 T cells, irrespective of CMV-

Gini-TCR skewing index in T cell differentiation subsets
To further validate our method, we determined Gini-TCR skewing index values in well-defined T cell differentiation subsets. T cells were therefore divided into naive (T Naive ), central memory (T CM ), effector memory (T EM ) and terminally differentiated (T TD ) T cells based on CD45RA and CCR7 expression [17] as shown in S3 Fig TCR sequencing and spectratyping studies have shown that naive T cells demonstrate the broadest TCR repertoire, whereas the repertoire becomes more contracted in memory T cells [18][19][20]. We indeed observed that CD4 T Naive cells show significantly lower Gini-TCR skewing index values than the three CD4 memory cells subsets (Fig 3A). Among the memory cells, CD4 T CM cells demonstrated significantly less TCR skewing than CD4 T EM and T TD cells, as previously found with TCR sequencing [18]. Further delineation of CD4 T Naive cells based on CD31 expression (S3 Fig) confirmed the previous observation that CD31 negative T cells (central naive) have a slightly more skewed TCR repertoire than CD31 positive T cells (thymic naive) [21], as demonstrated in Fig 3B. Analysis of TCR-Vβ distribution in CD8 differentiation subsets (S3 Fig) resulted in similar findings as observed in CD4 T cells, although the TCR repertoire skewing was more pronounced in the CD8 T EM and T TD populations (Fig 3C). Taken together, calculation of the Gini-TCR skewing index in naive and memory T cell subsets provided similar results as previously reported with TCR sequencing and spectratyping [18][19][20][21]. Hierarchical clustering analysis of Gini-TCR skewing index values As the distribution of the TCR repertoire was simultaneously measured in a large number of distinct CD4 and CD8 T cell subsets, we next determined to what extent TCR skewing in one subset is linked to TCR skewing in another subset. Hierarchical clustering analysis readily revealed that CD8 T EM and CD8 T TD cells formed a separate cluster from all other CD4 and CD8 T cell subsets (Fig 3D). This finding indicates that TCR skewing in CD8 T EM and T TD cells occurs relatively independently from TCR skewing in the other subsets. An additional hierarchical clustering analysis showed that Gini-TCR skewing index values of CMV-seropositive and CMV-seronegative individuals largely formed two separate clusters. This finding further underscores the impact that CMV has on the TCR repertoire in humans [14].
Tracking Gini-TCR skewing index changes following an antigenic challenge Finally, we determined whether the Gini-TCR skewing index allowed us to detect TCR repertoire changes following an antigenic challenge. Therefore, we studied the TCR repertoire of CD4 and CD8 T Naive and T EM cells in 9-year old children receiving a combination vaccine against diphtheria, tetanus, acellular pertussis and inactivated poliovirus (DTaP-IPV). We hypothesized that DTaP-IPV vaccination will recruit naive T cells towards the effector memory T cell pool and thereby cause skewing of the naive TCR repertoire. In accordance with our expectations, we observed clear skewing of the TCR repertoire among CD8 T Naive cells at 28 days after vaccination (Fig 4A, left panel). In parallel, we observed a slight decrease in Gini-TCR skewing index values, i.e. broadening of the TCR repertoire, among CD8 T EM cells at 28 days ( Fig 4A, right panel). The latter finding was only temporary, as Gini-TCR skewing index values of CD8 T EM cells returned to baseline levels at 1 year after vaccination. In contrast, Gini-TCR skewing index values remained elevated among CD8 T Naive cells after 1 year follow-up. These TCR repertoire changes were associated with a stable numerical decrease in CD8 T Naive cells (Fig 4B, left panel) and increase in CD8 T EM cells (Fig 4B, right panel). In essence, we observed comparable modulations within the CD4 T cell compartment (Fig 4C and 4D), albeit substantially less pronounced. These findings indeed implied that naive T cells are recruited towards the effector memory pool upon DTaP-IPV vaccination. Taken together, our results show that the Gini-TCR skewing index provides stable values over time, but is sensitive enough to detect TCR repertoire changes following an antigenic challenge.

Discussion
We here report a novel application of the Gini index for analyzing flow cytometric data on the TCR-Vβ repertoire. Flow cytometry is widely used to investigate the human TCR repertoire, as it quickly provides insight into the TCR repertoire of many different T cell subsets without the need for laborious cell-sorting. Until now however, the analysis and interpretation of the flow cytometric data on TCR-Vβ usage was difficult and subtle TCR repertoire changes were easily missed [6]. In the current study, we show that economic statistics, which have previously been applied in TCR sequencing studies [10,11], allowed us to quantify the flow cytometric data on proportional TCR-Vβ usage into direct measures of TCR-Vβ distribution: the Gini-TCR skewing index. Using the Gini-TCR skewing index, we were capable of detecting subtle changes of the TCR repertoire among multiple, well-defined populations of T cells. We validated the Gini-TCR skewing index by comparing our results to previous findings obtained with TCR sequencing and spectratyping [14,[18][19][20][21]. We confirmed that chronic CMV infection and aging are associated with substantial skewing of the TCR repertoire in CD8 T cells [14][15][16]. Moreover, we found that naive and memory T cell populations display differences in TCR repertoire skewing. Naive T cells are known to have the broadest TCR repertoire, whereas the repertoire becomes skewed in memory T cells [18][19][20]. By calculating the Gini-TCR skewing index, we could confirm that naive T cells have a less skewed TCR-Vβ repertoire than memory T cells. In addition, we verified that CD31+ thymic naive CD4+ T cells have a broader TCR-Vβ repertoire than CD31-central naive CD4+ T cells [21]. Furthermore, we confirmed that central memory CD4 and CD8 T cells demonstrate less TCR repertoire skewing than effector memory and terminally differentiated memory T cells [18]. Interestingly, we could even detect changes of the TCR-Vβ repertoire following an antigenic challenge (i.e. DTaP-IPV vaccination). Our findings therefore confirm that distinct populations of T cells are characterized by distinct skewing of the TCR repertoire, and that these differences can be detected with the Gini-TCR skewing index. Furthermore, our data underscore that future TCR repertoire studies should take into account that age, antigenic challenges and the presence of chronic viral infections (i.e. CMV) have a profound impact on the TCR repertoire.
In summary, we introduce a novel and rapid approach for analyzing flow cytometric data on TCR-Vβ usage. By applying economic statistics, we could calculate the Gini-TCR skewing index and readily track subtle changes of the TCR-Vβ repertoire among a large number of T cell subsets. Our study therefore reveals interesting synergy between economic statistics and the flow cytometric analysis of the TCR repertoire. We expect that application of the Gini-TCR skewing index to the flow cytometric TCR-Vβ analysis will contribute to a better understanding of the TCR repertoire in health and disease. The development of novel monoclonal antibodies for additional TCR-Vβ families could make the flow cytometric assessment of the TCR repertoire also more suitable for use in clinical diagnostics [4,5,22].