Immune repertoire is a term that is commonly used in immunology to describe the level of diversity and clonality of B and T cell antigen receptors, the immunoglobulins (IG) or antibodies and T cell receptors (TR). These cells encode an humongous variety of receptors that are capable of recognizing any organic macromolecule of biological relevance. The main process for the generation of the antigen receptors is called receptor rearrangement and is very similar for B and T cells: Every antigen receptor consists of two different chains that are responsible for antigen recognition, namely the α (TRA) and β (TRB) chain, and γ (TRG) and δ (TRD) for α β and γ δ TR, the immunoglobulin heavy chain (IGH), and one of two different immunoglobulin light chains (IGK, IGL) for the immunoglobulins or antibodies. IGH and TRB V domains are encoded by three different gene segments: variable (V), diversity (D) and joining (j); IGK, IGL and TRA V domains are encoded by two gene types, V and J [1]. A human genome in germline confirmation comprises alleles for every gene [2]. During B and T cell development the cells rearrange the genes so that there is only one V gene and one J gene per rearrangement (and usually one D for IGH and TRB, but several for TRD), and J element per functional exon. An important principle called allelic exclusion ensures that only one receptor specificity is expressed per B or T cell.

The human adaptive immune system has a strong impact on human health. Its efficiency is fundamentally reliant upon antigen receptor diversity; a restricted repertoire is in many cases unable to recognize the full variety of pathogens. In addition, an immune response as well as certain diseases lead to clonal expansions of B and T cells depending on their receptor specificity. Therefore, analyzing and understanding the repertoire is highly beneficial for research issues as well as to optimize medical treatment of patients [3].

Today’s most advanced techniques in immune repertoire analysis are based on next-generation sequencing (NGS) [4] that produces huge amounts of data. Currently, there exist various analysis and visualization tools for system immunology with different focuses such as, for example, MiTCR [5], Decombinator [6], IMGT/HighV-QUEST [7], IgBLAST [8], ImmunTraCkeR [9], immunoSEQ [10], IgAT Tool [11], and IgTree [12].

Some of those tools are focused on calculating a wide range of statistics (e.g., IgAT), performing alignments to facilitate analysis of the immunglobulin variable domain sequences (e.g., IgBLAST) or generating lineage trees from immunoglobulin variable region gene sequences (e.g., IgTree). All those tools are based on analyzing the B cell repertoire, while others enable detailed research on the T cell repertoire: For example, ImmunTraCkeR determines V-J rearrangements and sets the main focus on the cell immune repertoire diversity. MiTCR offers a fast CDR3 algorithm and a PCR two-stage approach for correcting sequencing errors. ImmunoSEQ mainly places emphasis on statistical analysis and visualization of IG and TR data.

Whereas most of these tools/frameworks are focused on one cell type or on one specific type of analysis, our here presented framework IMEX has been designed for comprehensive, in-depth analysis of human antigen IG and TR repertoires based on NGS data. IMEX contains algorithms for gaining more knowledge about the diversity on different sequence levels based on IMGT/HighV-QUEST analysis outputs [7, 13]. In the context of the calculation of clonality, IMEX users are able to define how to calculate sequence clonality and to compare diversity and clonality of various samples. A primer efficiency analysis enables the investigation of primer matching frequencies in PCR experiments. IMEX also includes V-(D)-J gene combination algorithms and additionally offers a wide range of visualization methods for gaining essential insights in the human adaptive immune system.

IMEX includes algorithms and statistical analyses for determining descriptive statistics about sequence functionality and V-(D)-J rearranged region frequency, calculating clonality of cells, estimating diversity of the cell spectrum, and visual representation of various gene/allele combinations. IMEX has been designed for analyzing and summarizing NGS-based IG and TR data derived from IMGT®;. IMGT/HighV-QUEST is a NGS high-throughput analysis portal for IG and TR, and so far the only one available online [7, 13]. IMGT/HighV-QUEST uses the same algorithms as IMGT/V-QUEST [14] with integrated IMGT/JunctionAnalysis [15], provides 11 compressed output files that contain information about variable (V), diverse (D), and joining (J) gene arrangements (V-(D)-J), identification and characterization of new alleles, detailed analysis of the junction (IMGT/JunctionAnalysis results), and additional information of mutations. IMEX uses these processed files as input for statistical analyses. Sample comparisons, clonotype tracking, and variety analysis are also included in IMEX. IMEX is written in C# and is freely available at http://bioinformatics.fh-hagenberg.at/immunexplorer/. In the following paragraphs we give detailed descriptions of the analysis methods implemented in IMEX.

Descriptive statistic analyses

IMEX enables a wide range of statistical analyses of IG and TR data. Lists of V, D, and J gene occurrences containing the total amounts and relative frequencies of these genes are calculated as well as the total amounts of the productive, unproductive, and unknown sequences (see Fig. 1). Sequences, for which no alignment result was found, are reported, but not considered later when it comes to further calculations in IMEX. Additionally, pie charts can be generated to gain more insights about the productive and unproductive B and T cell arrangements of the human adaptive immune system. All statistical calculations can be downloaded as text files and used for further calculations.

Diversity analysis

The diversity of an antigen receptor repertoire is calculated by analyzing the unique clonotypes of IG and TR in all sequences.

In the literature, several different ways to define the term diversity can be found [18]; IgAT, for example, calculates the clonotypic diversity as clonotypes per productive sequences and the sequence diversity as unique sequences per productive sequences [11]. IMEX calculates sequence diversity using a more elaborated data mining approach [19] based on the most variable region, the CDR3 [7]:

To empirically calculate the diversity in IG or TR data, we randomly choose n out of N CDR3 sequences (r a n d(n,N)) in the sample and determine the number of unique clonotypes (c _unique ) in these n sequences. This c _unique (n) is calculated for increasing numbers of n, for example for n={0,1000,2000,3000,…}, and so we get the calculated diversity d i v _calc (n) in n sequences:

$$\begin{array}{@{}rcl@{}} div_{calc}(n)= c_{unique}(rand(n,N)) \end{array} $$

(1)

This calculation is repeated five times for each n and the number of unique clonotypes c _unique is averaged. Examples are shown in Fig. 2.

We assume that there is a certain amount of unique clonotypes in the sample, and the more amino acid sequences we draw from the sample, the more the number of unique sequences will converge to the true number of unique clonotypes. Additionally, we have to keep in mind that the more sequences we draw, the more unique sequences we will see due to read errors. This is why we assume that the number of unique sequences (seen in n randomly drawn sequences) can be modeled as

$$\begin{array}{@{}rcl@{}} div_{mod}(n)= a * (1-e^{-b*n})+k*n \end{array} $$

(2)

where a is the true number of unique clonotypes and k is the fraction of unique sequences caused by read errors.

The parameters a, b, and k of the here proposed model are optimized so that they fit the empirically calculated diversity d i v _calc using evolution strategies [20]. The so optimized a in the model corresponds to the total number of unique clonotypes in the multiplex PCR as shown in Fig. 3.

Here we demonstrate the analysis of NGS data of a proband whose immune spectrum showed highly abundant clonal Expansion over a longer time period. Using analysis methods provided by IMEX we found two cytotoxic T cell clonotypes (CD8+) that are highly abundant and can be constantly observed over several months. The data sets have been obtained using PCR (Biomed 2 primer panels for gDNA amplification) of the IGH and TRB loci [21] followed by next-generation sequencing (Illumina Miseq sequencer).

We took blood samples of the proband p78690 at three different time points (November 2013 (T1), February 2014 (T2), and May 2014 (T3)); for every time point we generated three data sets, one of the IGH chain and two of the TRBV chain (primer sets 1 and 2). After having analyzed the data using IMGT/HighV-QUEST online (http://www.imgt.org), we performed statistical sequence analysis of the so generated data sets, the results are given in Tables 1 and 2. These data form the basis of a first, general overview of the IG and TR repertoires, shown in Fig. 4.

Proband: p78690	TRB primer set 1			TRB primer set 2			TRB primer set 2:	TRB primer set 2:
							CD4-/CD8+	CD4+/CD8-
	T1	T2	T3	T1	T2	T3	T3	T3
Productive	16,707	104,346	91,840	77,380	77,997	148,676	224,397	170,696
Unproductive	4018	22,534	18,739	9382	16,066	27,577	42,477	32,884
Unknown	4461	2639	1994	1424	67	1102	1403	1167
No result	11,989	10,879	9068	14,268	1162	14,694	11,498	8135
Total number of sequences	37,175	140,398	121,641	102,454	95,292	192,049	279,775	212,882

For each TR primer set (T cell receptor β chain amplification based on Biomed-2 primer sets) we have determined the total number of productive, unproductive, unknown and no result sequences. For primer set 2 at time point 3 we prepared an additional basic analysis for CD4-/CD8+ and CD4+/CD8- sorted cells

Proband: p78690	IGH
	T1	T2	T3
Productive	196,479	105,767	201,582
Unproductive	24,770	15,513	27,748
Unknown	1105	314	1769
No result	14,915	156	620
Total number of sequences	237,269	121,750	231,719

We additionally tested the contribution of the multiplex primers to the total number of generated sequences by using the IMEX PCR primer matching algorithm for quality control of the PCR; the results of this analysis are shown in Fig. 5. There we see that no amplicons derived from primer IGHV7 subgroup and almost no amplicons from primer IGHV6 subgroup are found in the sample.

In order to determine the variability of the IGH and TRB repertoire we analyzed and compared the V-(D)-J combinations of three different time points. As shown in Fig. 6, the TRB V-(D)-J rearrangement profile does not change over time, which means that the proband had no serious gene arrangement changes. We also see that there are two highly expanded V-(D)-J clonotypes that have to be analyzed in detail on gene level. Surprisingly we also found two highly abundant TRB CDR3s (AA) (ASSVSGEGSDEQF and ASSMGQNNEQF) for all three time points (see Fig. 7).