1446
BRIEF COMMUNICATIONS
2
vary significantly over colony (Kruskal-Wallis:
ϭ
4197.997, n 9, P Ͻ 0.0001). A low coefficient of variation
ϭ
indicates a low level of specificity in the response toward C.
bombi (little variation in the response toward different strains;
i.e., all strains can infect to a similar degree). A high coef-
ficient of variation, in contrast, indicates a high specificity
in the response to different strains (i.e., the response towards
strains varies greatly; in the present case, only a few strains
can infect). Crithidia, of course, would be fought by the non-
specific (general) immune response as well. However, any
effect due to the general immune response would thus be
equal for all strains in a single colony and would be ignored
by the coefficient of variation (which measures the relative
difference).
For each colony we also measured the non-specific immune
response, that is, the average encapsulation response of its
workers toward the general insult, the nylon implant. Similar
to the specific response, we found that encapsulation varies
significantly between colonies (data natural log(
ϩ
1) trans-
F
IG. 1. The strength of the nonspecific response (degree of en-
formed: one-way analysis of variance: F10, 32
ϭ
2.249, P
ϭ
capsulation; units) in relation to the susceptibility against an ex-
perimental range of different Crithidia bombi strains (measured as
the coefficient of variation, CoV, of each colony’s strain-specific
cell counts, for infection intensity; high values indicate high degrees
of specificity; i.e. exclusion of most parasites). Each point is the
0.04).
When comparing these two measures for the specific and
nonspecific arm of the immune system, respectively, we
found that the strength of the nonspecific response correlates
negatively with the ability to specifically resist a range of
different parasite strains (Fig. 1; Spearman’s rank correlation
mean value for all tested workers of a given colony (N
ഠ 20 in-
dividuals per colony; see text). The relationship is for coefficient
of variation versus encapsulation: Spearman’s rank r ϭ Ϫ0.767,
n
ϭ 9 colonies, P ϭ 0.016. The insets show the cell counts for that
r
ϭ Ϫ0.767, n ϭ 9, P ϭ 0.016). We also calculated two other
particular colony, each bar representing the percentage contribution
of each of the four strains to the total observed infection intensity.
The left inset is an example of an evenly distributed susceptibility
to the four parasite strains; that is, low specificity. The right inset
illustrates a specific response with an unevenly distributed suscep-
tibility. The error bars represent the standard errors of the means
calculated for the x-axis by bootstrapping the original population.
measures of variation (Zar 1996), Simpson’s diversity index
and Berger Parker’s index in the place of the coefficient of
variation, but obtained similar results (vs. encaps: Simpson’s
r
ϭ 0.761, n ϭ 9, P ϭ 0.016; Berger-Parker r ϭ 0.961, P Ͻ
0.0001). This shows the overall result is not due to our choice
of variation measure.
terrestris, see Ko¨nig and Schmid-Hempel 1995; Moret and
Schmid-Hempel 2000) and there is a cost of virulence for
the parasite. In his analysis, Frank found that hosts well pro-
tected by an array of specific responses should be selected
to reduce the investment for the nonspecific immune response
(and vice versa, in different circumstances, for high levels
of nonspecific immunity).
It is important here to emphasize that we are not sug-
gesting, for example, that highly specific colonies are fitter
than those with a strong general immune arm, or vice versa.
Rather the aim of the study was to elucidate the particular
combination of specific and general immune abilities. Sim-
ilarly, at present it is not known what kind of physiological
mechanisms might be responsible for the observed inverse
relationship (Fig. 1). Our experiments only suggest that hosts
may not be capable of maintaining strong nonspecific re-
sponses and specifically resist many different strains of a
parasite at the same time. Despite the incapacity to suggest
any particular mechanism, the results shed a novel light on
the possible constraints that affect the evolution of the im-
mune system and particularly on the price that is paid for
specificity. In fact, because we have measured the respective
responses separately, in different individuals (but from the
same host line), the negative relationship is not tied to the
D
ISCUSSION
We set out to discover the relationship between the strength
of the nonspecific immune response and the specificity of the
specific immune response. Despite a relatively restricted sam-
ple for each within-colony and strain assay (imposed by the
biological constraints of the system), we found that workers
of a colony, with a good ability to defend themselves against
parasites requiring a specific response, respond only weakly
to a general insult (such as encapsulation of a novel antigen).
Natural selection may operate so that good quality hosts
show a strong nonspecific encapsulation response and at the
same time resist a wide range of parasite strains (i.e., are
highly specific). Alternatively, the capacity to mount a highly
specific response against many different parasite strains could
compromise or even negate the need for the ability to deal
with a generalized insult. Here we have evidence for the latter
(Fig. 1). The results are consistent with predictions from
theoretical models (Frank 2000) that seek to understand the
combination of factors that favor either specific or nonspecific
responses. The Frank model is a simple Lotka-Volterra sys-
tem, in which hosts must compete for limited resources. Spe-
cific defense in the model is based on each host carrying a
specific recognition allele; specific defense preventing an at-
tack when the host shares the corresponding allele. Both types simultaneous activation of different components of the im-
of defense carry costs for the host (as is the case for B. mune system within the same individual (that is, likely does