Clusters versus affinity-based approaches in F. tularensis whole genome search of CTL epitopes

Abstract

Deciphering the cellular immunome of a bacterial pathogen is challenging due to the enormous number of putative peptidic determinants. State-of-the-art prediction methods developed in recent years enable to significantly reduce the number of peptides to be screened, yet the number of remaining candidates for experimental evaluation is still in the range of ten-thousands, even for a limited coverage of MHC alleles. We have recently established a resource-efficient approach for down selection of candidates and enrichment of true positives, based on selection of predicted MHC binders located in high density "hotspots" of putative epitopes. This cluster-based approach was applied to an unbiased, whole genome search of Francisella tularensis CTL epitopes and was shown to yield a 17-25 fold higher level of responders as compared to randomly selected predicted epitopes tested in Kb/Db C57BL/6 mice. In the present study, we further evaluate the cluster-based approach (down to a lower density range) and compare this approach to the classical affinity-based approach by testing putative CTL epitopes with predicted IC(50) values of <10 nM. We demonstrate that while the percent of responders achieved by both approaches is similar, the profile of responders is different, and the predicted binding affinity of most responders in the cluster-based approach is relatively low (geometric mean of 170 nM), rendering the two approaches complimentary. The cluster-based approach is further validated in BALB/c F. tularensis immunized mice belonging to another allelic restriction (Kd/Dd) group. To date, the cluster-based approach yielded over 200 novel F. tularensis peptides eliciting a cellular response, all were verified as MHC class I binders, thereby substantially increasing the F. tularensis dataset of known CTL epitopes. The generality and power of the high density cluster-based approach suggest that it can be a valuable tool for identification of novel CTLs in proteomes of other bacterial pathogens.

Figure 1. Flowchart of down-selection of putative MHC binders for experimental evaluation.

The whole-genome immunoinformatic analysis conducted on the 1754 F. tularensis ORF products and the downstream selection of the various subsets of peptides is provided. Boxes shaded in light blue: Subsets I and II were described in a previous study and are added for the sake of completeness. Boxes shaded in cyan: Subsets III–V of peptides selected and evaluated in this study. In all down selection steps referred to as : “Selected for evaluation" (in Subset III, IV and V), clusters or peptides are selected at random as an operational filter for reduction of number of peptides to be experimentally tested, unlike Subset II, were selection of peptides was at random with respect to clusters. Subset III contains 401 peptides selected from clusters having density of 0.8 up to 1.0 (1.0 not included); Subset IV is composed of 200 peptides selected from the overall 3016 predicted MHC binders (IC₅₀< = 500 nM) located outside the high-density cluster regions in the 41 source proteins of Subset I (Figure 2); Subset V includes 92 predicted epitopes having an IC₅₀ of 10 nM and lower, representing putative high-affinity MHC binders.

Figure 2. Array of peptides tested in the F. tularensis CTL screen in C57BL/6 LVS immunized mice.

Summary of the total number of peptides selected, evaluated and responding in the various subsets (see Figure 1). Data on Subset I and II is from . Peptides which were shown to stimulate lymphocytes and induce IFNγ production are referred to as “responders". P-values are provided for evaluation of significance of difference between number of responders in the cluster-based Subsets I and III versus Subset II, and versus Subset IV.

Figure 3. Magnitude of T-cell response among identified CTL epitopes (Subsets I, III, V).

Responders identified from clusters with densities of 1.0–1.4 (dark cyan) (determined from [14]), from clusters with densities of 0.8–1.0 (light cyan) and from the high affinity-based approach (light blue) are classified according to magnitude of T-cell response, as deduced from number of spots in the EliSpot assay. The bars represent the percent of responders per subset in each of the classes, while the actual number of responders is given in brackets on the top of the corresponding bar. Classification (expressed in number of spots/million cells) is: Low: 5–20; Medium: 20–32; High> = 33).

Figure 4. Distribution of parental proteins containing responders according to their membrane topology.

Parental proteins containing responder peptides identified by the cluster-based, high affinity-based and random-based approaches are classified as follows: Membrane-spanning proteins (with 10 or more predicted transmembranal domains (TM) and/or TMs spanning at least a third of the protein) in dark blue; partially membranal proteins (less than 10 predicted TMs) in blue; proteins with no predicted transmembranal domains (except signal peptide) in light blue. . P-values are provided for evaluation of significance of difference between membranal proteins in the cluster-based Subsets I and III versus Subset V and versus Subset II.

Figure 5. Effect of anti-CD8 and anti-CD4 antibodies on the T-cell response.

A sample of over 20 EliSpot plates is presented, in which each of the first ten columns contains splenocytes incubated with an individual peptide and the last two columns are positive and negative controls, containing splenocytes incubated with formalin killed LVS (LVS) and splenocytes devoid of antigen (No Ag), respectively. Rows 1 and 2 (duplicates) contain splenocytes incubated with peptides denoted 001-010 (arbitrary numbering), or with LVS; the following pairs of rows (from top to bottom) are repeats of rows 1 and 2 supplemented with anti-CD8, anti-CD4 or a mixture of anti-CD4 and anti-CD8 antibodies. All 222 F. tularensis responder peptides exhibited the same behavior as depicted in the sample plate shown in this figure, with respect to the effect of anti-CD8 and anti-CD4 antibodies (see text and (Table S5)).

Figure 6. Distribution of clusters of CTL epitopes in a representative non-membranal and membrane-spanning protein and alignment with transmembrane α-helices and loops.

Sample plots of the topological map (generated by TMHMM predictions) of two representative proteins. Red segments represent the probability of having a helical region, while the thin blue and pink lines represent inside and outside loops, respectively. In the lower part of every protein chart, a bar represents a cluster of predicted epitopes. Co-localization of the putative MHC binders with the predicted helix or loop region in the protein is marked by a grey box. (A) A sample of a non-membranal protein (gi|89256893) carrying a cluster at density 0.8, harboring 23 putative epitopes, four of which are a responders. This represents the 18% non-membranal proteins carrying high density clusters (see Figure 4). (B) A sample of a transmembrane protein (gi|89256946) carrying four clusters (clusters #1–4) within a α-helical transmembrane domain, (marked by “TM"), and an additional cluster (cluster #5) co-localized with an outside loop domain (marked by “loop") between two transmembranal regions. The total number of putative epitopes in clusters #1–4 is 102 and in cluster #5 is 17, which contain 3 and 2 responders, respectively. This sample protein provides an example for the high frequency of clusters in helical regions but also demonstrates the presence of high density clusters in loops (there are examples of other proteins, not shown, where clusters span both transmembrane and loop regions).