Bovine tuberculosis

178 R. Waters and M. Vordermeier

CFP-10 and Rv3615c reagents (Jones and
Vordermeier, unpublished data).

12.5.3 Non-biased genome-wide

approaches

We have over the past 15 years assessed the
immunogenicity of 626 M. bovis/M. tuberculosis-
derived proteins, either in the form of
recombinant proteins or, more frequently, as
overlapping sets of synthetic peptides. The high-
throughput readout system to generate these
data was the blood-based IGRA introduced in
previous sections of this review. This analysis
allowed us to define antigenicity in infected cat-
tle based on the frequency a particular antigen
was recognized (responder frequency). As
shown in Fig. 12.1, a response hierarchy could
be established in this manner ranging from no
recognition (responder frequency = 0%) to fre-
quencies of around 90% (Fig. 12.1). When the
results were stratified according to protein func-
tionality or membership of particular protein
families, we could confirm the dominance of
ESAT-6-family members, predicted secreted
proteins and members of the PE/PPE families
( Vordermeier and Jones, unpublished data).
However, the studies, while contributing to pop-
ulate our response hierarchy (Fig. 12.1), as well

as earlier and more empirical approaches, were

not unbiased as they depended on an underlying

assumption on what rendered a protein immu-

nogenic, such as being secreted, highly expressed

or being a PE or PPE protein. Therefore, it is

to date not possible to conclude fully that

particular functionalities of proteins can be

related to their immunogenicity. Thus, unbiased,

proteome- wide immunogenicity mapping to

define the T-cell antigenome of M. bovis in cattle

would be highly beneficial to provide a platform

to rationally explore the specificity of immuno-

genic proteins.

An unbiased approach taken towards this

goal was to prepare proteins based on a Gateway

library from M. tuberculosis and to screen these

proteins with blood from infected or uninfected

cattle using IGRA (Jones et al., 2013). In this

approach, however, protein quality (e.g. purity)

had to be de-prioritized in favour of quantity.

While this study lead to the identification of

potential new subunit vaccine candidates

(Jones et al., 2013), it also highlighted several

limitations of this approach, namely, that due

to low purity, the cut-off for classifying a posi-

tive response had to be increased to mitigate

against false-positive responses, which in turn

most likely resulted in potential antigens being

missed out due to the decreased sensitivity of

the readout system. This system was also not

able to produce a library of proteins that covered

the whole proteome within a reasonable period

and protein amounts required for T-cell

screening.

Recently, an alternative approach was

applied to human TB, namely the combination

of robust computational methods to predict

proteome- wide peptides binding to human HLA

molecules combined with high throughput pep-

tide synthesis and ELISPOT-based T-cell assays.

The study by the group of Sette (Lindestam

Arlehamn et al., 2013) established an immuno-

logical footprint of M. tuberculosis CD4 T-cell rec-

ognition, thereby demonstrating that CXCR3+/

CCR6+ memory T cells are highly focused in

their recognition towards three immunodomi-

nant antigenic islands mapping to ESAT-6 fam-

ily members associated esx secretion systems

(Lindestam Arlehamn et al., 2013). To imple-

ment a similar strategy for cattle depends on the

availability of similarly robust bovine MHC class

Fig. 12.1. Hierarchy of T-cell responses to 626
M. bovis/M. tuberculosis proteins. Results are
shown as responder frequencies (proportion of
tested animals responding to a given protein).
Responses were established using whole blood
cultures from M. bovis-infected cattle to measure
antigen-specific IFN-γ responses.

100

1

101 201 301 401 501 601

75

Proteins

Responder frequencies (%)

50

25

0