conspecific ant aggression is correlated with chemical distance, but not with genetic or spatial...
TRANSCRIPT
ORIGINAL PAPER
Conspecific Ant Aggression is Correlated with Chemical Distance,but not with Genetic or Spatial Distance
Stephen J. Martin • Emma Vitikainen •
Falko P. Drijfhout • Duncan Jackson
Received: 21 April 2011 / Accepted: 8 September 2011 / Published online: 18 September 2011
� Springer Science+Business Media, LLC 2011
Abstract Five possible mechanisms might underlie kin
recognition in social groups: spatial location, familiarity
through prior association, phenotype matching, recognition
alleles, or rejecting unfamiliar cues. Kin recognition by phe-
notype matching relies on a strong correlation between
genotype and phenotype. Aggression bioassays are the
standard method for investigating recognition in animals,
particularly social insect interactions among nestmates and
non-nestmates. These bioassays typically pay little regard to
how outcomes are determined by differences in chemical
recognition cues of the test subjects, because the system of
signal coding was unknown until recently. We exploited the
known nestmate recognition system of the ant Formica
exsecta to investigate aggression between 24 pairs of colonies
across a range of chemical (Z9-alkene & n-alkanes), genetic,
and spatial distances. The whole Z9-alkene chemical profile
was the only significant (p \ 0.001) predictor of aggression
levels. Aggression was a nonlinear step function of Z9-alkene
chemical distance, where a small change in chemical profile
resulted in a rapid behavioural transition from non-aggres-
sion to overt aggression. These findings raise questions
surrounding our current understanding of recognition systems,
because they support phenotype matching to a colony chem-
ical profile without a significant genetic or spatial component.
Keywords Alkenes � Aggression � Formica exsecta �Kin recognition
Introduction
The ability to discriminate friends from foes is a funda-
mental component of basic behaviours, such as mating,
aggression and aggregation (Wyatt 2003). Five possible
mechanisms might underlie recognition in social groups:
spatial location (Beecher et al. 1981), familiarity through
prior association (Reeve 1989), phenotype matching (Mateo
2009), recognition alleles (Keller and Ross 1998), or
rejecting unfamiliar cues (Ozaki et al. 2005; Guerrieri et al.
2009). Evidence from vertebrates and insects favours phe-
notype matching mechanisms that allow recognition irre-
spective of spatial location, whereas, allele recognition
(genetic kin recognition) has little support. The recognition
phenotype should hypothetically be correlated closely with
genotype in social groups, to ensure that altruism benefits
close kin, thereby increasing indirect fitness.
The phenotype matching model is commonly used for
understanding recognition, where the mechanism for
matching might involve reference to a learned phenotypic
template or self-reference (Lacy and Sherman 1983), a
mixture of both (Newey 2011), or rejecting unfamiliar cues
(Ozaki et al. 2005; Guerrieri et al. 2009). Presumably,
recognition cues are detected and compared with a neural
template and if the cues do not match the familiar template
then an aggressive reaction is elicited. Thus, aggression
serves as one of the best yet available proxy for measuring
Edited by Bambos Kyriacou.
S. J. Martin (&) � D. Jackson
Department of Animal & Plant Sciences, University of Sheffield,
Sheffield S10 2TN, UK
e-mail: [email protected]
E. Vitikainen
Department of Biological and Environmental Sciences,
University of Helsinki, P.O. Box 65, 00014 Helsinki, Finland
F. P. Drijfhout
Chemical Ecology Group, School of Physical and Geographical
Sciences, Lennard-Jones Laboratory, Keele University,
Staffordshire ST5 5BG, UK
123
Behav Genet (2012) 42:323–331
DOI 10.1007/s10519-011-9503-0
the recognition abilities of individuals in many taxa.
Recognition is particularly important in social groups and
the best-studied examples are social insects and verte-
brates. Many different sensory modalities might be used for
social recognition by vertebrates, but odours have a pre-
dominant role in social primates and rodents, such as ring-
tailed lemurs (Palagi and Dapporto 2006) and Belding’s
ground squirrels (Mateo 2009). Similarly, the ability to
discriminate between nestmates and non-nestmates in
social insect colonies (bees, ants and wasps) uses chemical
recognition cues that ensure altruism is directed towards
nestmates (Holldobler and Wilson 2009). However, unlike
mammals (Palagi and Dapporto 2006) a strong genotypic
basis for recognition seems unlikely in social insects. This
is because a single colony often contains individuals from a
diversity of genetic backgrounds, due to multiple mating of
a single queen or the presence of multiple queens, and
highly accurate genetically-based kin recognition would
lead to nepotism, which would be detrimental to overall
colony functioning (Boomsma et al. 2003). Thus, previous
researchers have found limited evidence for genetic kin
recognition and hypothesized that if it persists then it might
only provide weak information that is fine-tuned to avoid
selfish exploitation (Boomsma et al. 2003; Nehring et al.
2011). However, the prevailing ‘gestalt hypothesis’ sug-
gests that individual recognition cues are mixed by chem-
ical transfer to produce an overall colony odour, which
eliminates any possibility of nepotism whilst ensuring a
high fidelity of colony recognition (Errard et al. 2006).
A part of our knowledge regarding the relative contri-
butions of environmental and genetic factors in recognition
is based on aggression studies. Despite this no consistent
method is used for selecting bioassay approaches, except
that ‘individual researchers tend to use the same assay
repeatedly’ (Roulston et al. 2003). A comparison of bio-
assay techniques found that methods differ greatly in their
ability to detect aggression, with the most significant fac-
tors being context and replicate number (Roulston et al.
2003). Recently mandible opening response has been used
in order to help standardise aggressive response of
restrained ants (Guerrieri and d’Ettorre 2008).
Aggressive interactions occur as a consequence of
individuals detecting chemical recognition cues that do not
conform to a familiar template (Ozaki et al. 2005; Guerrieri
et al. 2009). It is expected that any recognition system
cannot be error-free (Waldman 1988), simply because of
individual phenotypic diversity, so non-kin might occa-
sionally be wrongly accepted into a colony. An error-free
recognition system would need a template with very nar-
row and biologically improbable tolerances, thereby mak-
ing it prone to rejection of kin, which is always costly. An
intruder that closely matched a colony template might
initially gain access to a colony by passing the inspection
of one or two individuals, but would eventually be detected
after multiple interactions (Johnson et al. 2011).
Understanding the evolution of recognition demands a
measure of the extent to which a recognition cue must fail
to conform to a template before aggression ensues. In the
case of social insects, the difference between recognition
cues can be measured as a chemical distance between
individuals (Guerrieri and d’Ettorre 2008). Understanding
how chemical distance determines the likelihood of
aggressive interactions, and the function of that relation-
ship, will provide us with a powerful means of investi-
gating behaviours mediated by recognition, especially by
identifying the constraints of a chemical recognition sys-
tem. The recent identification of the exact encoding used
by a species for nestmate recognition mediated by the
Z9-alkene cuticular hydrocarbons (Martin et al. 2008a, b),
means we are now in a position to test the limits on rec-
ognition accuracy imposed by chemical distance.
In this study we employed the nestmate recognition
system of Formica exsecta ants (Martin et al. 2008a, b) to
investigate con-specific aggression between 24 pairs of
colonies, where we determined their chemical, genetic and
spatial distances.
Materials and methods
Study site and species
This study was conducted using F. exsecta colonies that form
part of the well-studied population found on the islands
Joskar, Furuskar and Rovholmen, at the Tvarminne zoo-
logical station in Hanko, Finland (Haag-Liautard et al. 2009;
Sundstrom et al. 2003; Vitikainen 2010). All nests were
monodomous (single mound), and the majority contain a
single queen (monogynous) with an average mating fre-
quency of 1.27 (Vitikainen 2010), but a small number
(n = 6) of monodomous colonies contained multiple queens
(polygynous). The population averages for within-colony
relatedness based on 8–24 workers per colony calculated
with Relatedness 5.0 (Queller and Goodnight 1997) were as
follows: polygynous colonies n = 7, r = 0.328; monogy-
nous polyandrous colonies n = 12, r = 0.622; monogynous
monoandrous colonies n = 68, r = 0.769. Preliminary
observations involved placing con-specific ants from both
monogynous and polygynous colonies onto the mounds of
polygynous colonies and observed host workers attacked the
non-nestmate ants indicating that polygynous colonies were
aggressive towards non-nestmates, so both colony structures
were included in the study. Genetic profiles, chemical pro-
files, and exact locations were available for all 80? active
colonies. We designed an aggression bioassay to test con-
specific nestmates versus non-nestmates, consisting of pairs
324 Behav Genet (2012) 42:323–331
123
of colonies with different combinations of genetic, chemical
and spatial distance. We split all colonies into groups
that were ‘close’ or ‘far’, with respect to their chemical
(Z9-alkenes), genetic (FST), or spatial distance (m), giving
eight possible combinations (23). We selected three pairs of
colonies for each of the eight possible combinations,
resulting in 24 pairs of colonies. These combinations utilised
33 of the 80 currently active colonies (Table 1).
Genetic distance
To determine the genetic structure of each colony 16
workers were genotyped at ten highly variable microsat-
ellite loci: Fe11, Fe13, Fe17, Fe37, Fe38, Fe42, Fe49
(Gyllenstrand et al. 2002), Fl21 (Chapuisat 1996), P22
(Trontti et al. 2003), and Fy3 (Hasegawa and Imai 2004),
following the method of Haag-Liautard et al. (2009). DNA
was extracted following the Chelex� protocol (Martin
et al. 2008b). PCR products were analyzed using a
MegaBACE 1000 sequencer with an ET400-R size stan-
dard. The genetic structure of the colonies was inferred as
the minimum number of parents needed to explain the
observed worker multilocus genotypes. We calculated the
pairwise fixation index (FST) for each pair of colonies
using the program Fstat 2.9.3.2 (Goudet 1995, 2001). The
resulting FST indices were used as a measure of genetic
distance between each pair of colonies, relative to the
overall population. So the larger the value the more
genetically different or ‘distance’ the pair of colonies are.
Chemical distance
The mean cuticular hydrocarbon (CHC) profile for each
F. exsecta colony in the population was determined in
September 2008 (unpublished data), by analyzing ten
individual workers per colony using gas chromatograph-
mass spectrometry (Martin et al. 2008a), and we identi-
fied a wide range of chemo-types (unpublished data).
Formica exsecta has a very simple CHC profile, consist-
ing of a homologous series of n-alkanes and Z9-alkenes,
where the individual peaks in either n-alkane or Z9-alkene
profiles were non-independent because they were the
products of specific biosynthetic pathways (Fig. 1). This
non-independence makes multivariate analysis unsuitable
for the analysis of profile data. Each F. exsecta colony has
a specific Z9-alkene profile (Martin et al. 2008a, b) that
remains relatively stable over a period of at least 3 years
(unpublished data). The chemical distance (cd) was
defined as a measure of the magnitude of the chemical
differences we were investigating. We used the Z9-alkene
profile to calculate the chemical distance, which we
measured using a ‘city block’ design. The differences in
proportion (%) for each Z9-alkene between a pair of
colonies were summed, before dividing this value by two
(Fig. 2). The division accounts for the fact that the change
in one compound must be accompanied by an equal
change in proportions of the other compounds, because
these are compositional data. This method produces a
value that ranges between 0 (identical profiles) and 100
(profiles that do not share any compounds) so all mea-
surements using the city block design are displayed as
percentages. We found this method was more practical
than calculating the Mahalanobis distance, because it did
not require data transformations (e.g. log-contrast or Geo-
mean) that might distort the profile and potentially attri-
bute more power to less abundant compounds, relative to
the major compounds (Martin and Drijfhout 2009a) and
also avoids having to replace any zeros with a small
arbitrary value. Furthermore, measuring chemical distance
based only on qualitative differences (Guerrieri and
d’Ettorre 2008) is unsuitable, since colonies profiles only
differ quantitatively in this species. Pair combinations
were selected on the basis of their 2008 Z9-alkene pro-
files, but we collected five ants from each study colony in
September 2010 for CHC profile reanalysis. This ensured
that calculations of chemical distance reflected colony
Table 1 Eight combinations of the two groups, i.e. close (C) or far (F) with three distance types, i.e. chemical (Ch), genetic (G), and spatial (S)
Combination Genetic (FST) Chemical Z9-alkenes (city block) Spatial (m)
Ch–C, G–C, S–C 0.3520, 0.0621, 0.1935 4, 8, 3 5, 9, 14
Ch–C, G–C, S–F 0.1783, 0.2478, 0.2402 9, 5, 9 1029, 344, 318
Ch–C, G–F, S–C 0.4986, 0.4809, 0.4261 2, 3, 8 18, 18, 26
Ch–C, G–F, S–F 0.3378, 0.3798, 0.4162 11, 3, 9 87, 260, 275
Ch–F, G–C, S–C 0.2356, 0.2985, 0.3094 46, 54, 35 22, 29, 53
Ch–F, G–C, S–F 0.2120, 0.3465, 0.2861 51, 15, 38 110, 863, 470
Ch–F, G–F, S–C 0.4516, 0.3784, 0.4332 25, 51, 31 11, 60, 14
Ch–F, G–F, S–F 0.5418, 0.3780, 0.5366 20, 59, 57 254, 928, 185
Distances for each of the three measures are shown and their distributions are shown in Fig. 3. Three values are provided under each heading,
because each combination was replicated three times
Behav Genet (2012) 42:323–331 325
123
profiles at the same time aggression bioassays were con-
ducted. We also calculated n-alkane chemical distances
using 2010 data and tested for any correlation with
aggression, although n-alkanes were previously found to
be uninvolved in nestmate recognition (Martin et al.
2008a, b).
Spatial distance
The location of each F. exsecta colony was recorded via
GPS. We used this data to calculate the Euclidean distance
(m) between each colony pair, irrespective of whether
colonies were on different islands, since these islands have
all been colonised by queens that have flown directly over
the water to reach them.
Aggression bioassays
Preliminary observations established that dyadic encoun-
ters occurring in a neutral arena, such as a Petri dish, did
not result in aggression. Instead, non-nestmates were
placed on the mound of another colony, where aggression
was instantaneous. The intruder was usually grabbed by its
appendages before being dragged from the nest mound, or
pinned down by several ants. All 53 aggression bioassays
(24 bi-directional pairs ? five control experiments) were
conducted on mound surfaces in the field, during Septem-
ber 2010. These experiments were partially blinded,
because the 2010 chemical distances were unknown until
after the study, and genetic distance data was not available
in the field. The observer was aware of the spatial distance
between each pair. From each colony pair, a small frag-
ment containing ten ants was placed into a Fluon-coated
box. Each ant was marked on the abdomen with a small dot
of white paint then transported to the nest site of its paired
colony, where they were placed individually onto the test
mound. We observed interactions between the marked non-
resident ant and first five resident ants. Each interaction
was scored as either aggressive (grabbing of appendages or
dragging off mound), or non-aggressive (antennation or
trophallaxis). The marked ant was typically grabbed
immediately by one or more ants and not released until
it was either killed, or dragged from the mound, which
could take several minutes. When a marked ant was held
for longer than 1 min, recording was stopped and any
remaining interactions assumed to be aggressive, because
other ants normally joined in the attack on the marked ant.
If the marked ant was released within a minute, then the
next encounter was recorded until either five interactions
were completed, or the ant was held for more than 1 min.
Elution order
C23
C25
C27
C29C23:1
C27:1
C25:1
C29:1
Ion
abun
danc
e
Fig. 1 Total ion chromatogram for a F. exsecta worker ant, showing
the very simple profile consisting of a C23 to C29 n-alkane and C23:1
to C29:1 Z9-alkene homologous series
0
10
20
30
40
50
60
1045
30
15
1043
32
15
0220
4/2 =2
Difference Difference
C23:1
C25:1
C27:1
C29:1
0
10
20
30
40
50
60
Colony C
1045
30
15
5530
105
4515
2010
90/2 =45
Colony AColony BColony A
Prop
ortio
n of
Z9-
alke
ne
Prop
ortio
n of
Z9-
alke
ne
Fig. 2 Calculation of chemical
distance (cd) between two
colonies with ‘close’ profiles
(left, cd = 2%) and ‘far’
profiles (right, cd = 45%),
using a city block design.
Differences between the same
peaks were calculated and the
direction of any difference
ignored, before all differences
were summed and the total
divided by two
326 Behav Genet (2012) 42:323–331
123
Each bioassay used ten marked ants, but if the first five
marked ants were all grabbed immediately then the
experiment was stopped. Since aggressive interactions
require physical contact between two ants they are always
very local and since the mounds are large relative to the
size of the ant the presence of a non-nest mate did not
appear to raise the level of excitement on the mound, since
no increase in aggression over time was seen towards non-
nest mate or control nest mate ants. Furthermore, since a
large number of host ants where present on the mound
meant that it was unlikely that the same host ants were
involved in repeated interactions with the non-nest mate
ants. The control bioassays involved precisely the same
procedure as above, except that the marked ants were
placed back onto their own mound and the first five
interactions recorded.
Data analysis
For each colony pair the proportion of aggressive interac-
tions was calculated by combining the data from both sets
of aggression bioassays. Simple linear regression was
performed to determine the significance of each variable as
a predictor of the aggression response. Segmented regres-
sion was performed only on Z9-alkene chemical distance
data as a predictor using the Segreg software package
(www.waterlog.info.segreg.htm), so as to test for the pos-
sibility of segmentation, or a step function in the response
that would indicate a response threshold. Interaction effects
were then tested using a multiple regression model in SPSS
v14, with a two step process, where the Z9-alkene chemical
distance was entered first (Step 1), because there is prior
evidence (Martin et al. 2008a, b) that this may be an
important predictor, before adding n-alkanes, genetic dis-
tances and spatial distances (Step 2).
Results
No aggression (prolonged fighting or dragging) was
observed during the 125 pairs of control interactions (five
nestmate interactions per ant 9 five ants per colony 9 five
colonies) between marked ants and their nestmates on the
test mounds. This indicates that neither paint marking, nor
removal/replacement of ants affected their behaviour. The
hydrocarbon profiles of nest mates from each colony were
very similar. Mean (SD) Z9-alkene intra-colony (nestmate)
distance was only 3 ± 1% for the 33 study colonies.
Aggression was always observed when non-nestmates
were placed on test mounds, although, the proportion of
aggressive encounters was highly variable ranging from 2
to 100% (Fig. 3). We found that the difference in aggres-
sion response was less than 10%, in 17 of the 24 (71%)
reciprocal pairing bioassays. However, we observed some
large differences (10–50%) between the remaining reci-
procal pair trials, especially when reciprocal pairs were
chemically similar, i.e. \10%. There was a significant
negative relationship between the difference in aggression
levels displayed by a reciprocal pair of colonies and their
Z9-alkene distance (Spearman’s rho, n = 24, r = 0.445,
p = 0.03). No significant relationship was detected for
genetic, spatial, and n-alkane distances, or variation in
response between the reciprocal pairs. A simple linear
regression analysis found that Z9-alkene chemical distance
was the only significant (p = 0.001, R2 = 0.414) predictor
of aggression (Fig. 3). Testing the relationship between
Z9-alkene chemical distance and aggression using linear
segmented regression found that the data were best fitted
by a step function with a break-point at a Z9-alkene
chemical difference of 9.41% (ANOVA, explained by
linear regression: F1,22 = 16.74, n = 24, p = 0.0005;
ANOVA, explained by breakpoint: F1,21 = 12.64, n = 24,
p = 0.0023; aggression (0 \ cd \ 9.41) = 4.10; aggres-
sion (cd [ 8.27) = 96.1). Thus, the actual biological
response function takes the form of a step function, where
inter-individual chemical differences exceeding a threshold
value (cd 9.41%) resulted in a very high probability of
aggression. Our simple analysis shows that aggression
levels were variable when the difference between colony
Z9-alkene profiles was low, but aggression was almost a
certainty when Z9-alkene profiles differed by more than
9.41% (Fig. 3). Multiple regression analysis (Table 2)
found that the addition of the other factors did not signif-
icantly (p = 0.35) improve the model. Therefore, the
n-alkane chemical distance, genetic distance, and spatial
distance were not significant predictors of aggression
(Table 2). This latter analysis also found that a quadratic
function of Z9-alkene chemical distance could predict
aggression with higher probability (p = 0.0001, R2 =
0.54), although, our step function is most biologically
valid. Thus, Z9-alkene chemical distance was the only
significant predictor of aggression. Further analysis of
chemical differences for each separate Z9-alkene showed
that, in contrast to the entire Z9-alkene blend, no individual
Z9-alkene was a significant predictor of aggression i.e.
C23:1 p = 0.8, C25:1 p = 0.75, C27:1 p = 0.66, C29:1
p = 0.95, C31:1 p = 0.47.
Discussion
In F. exsecta, the chemical distance between colony pro-
files for the Z9-alkenes was the only significant predictor of
aggression, explaining up to 54% of the variation. Chem-
ical distances of the n-alkane profile, genetic distances,
and spatial distances were neither significant predictors of
Behav Genet (2012) 42:323–331 327
123
aggression, nor did they increase the predictive power of
the Z9-alkene profile distance when tested in models of
interaction effects (Table 2, DR2 = 0.07). Reeve (1989)
suggested that familiarity, i.e. contact with neighbors,
might affect the recognition threshold, but we found no
spatial effect. Neither did we find aggression to be directly
influenced by levels of genetic relatedness. This adds to the
growing body of evidence that an increase in genetic
diversity does not increase cue diversity (Martin et al.
2009; Schmidt et al. 2010; Helantera et al. 2011).
Although, it has long been assumed that lower relatedness
correlates with greater genetic diversity in all genes, it is
possible that only a few genes involved in CHC production
could strongly impact on cue diversity, irrespective of
relatedness. However, this cannot be tested until we fully
understand the entire suite of genes that control CHC
production in insects. It is well established that the bio-
synthetic pathways of hydrocarbons such as n-alkanes and
Z9-alkenes are under genetic control (Morgan 2004;
Blomquist and Bagneres 2010), although, their expression
can be influenced by environmental factors (Wagner et al.
2001).
When non-nestmates encounter chemically distant resi-
dent ants on their nest mound they were generally attacked
by the first ant they encounter, but when their Z9-alkene
profiles were similar there were frequent errors and
aggression does not always occur on first contact. Behav-
ioral interactions between chemically close individuals
from different colonies ranged from trophallaxis (food
transfer), to antennation, and ultimately attack, although
the number of encounters needed before attack was highly
variable. This suggests that there is individual variation in
the threshold for recognition of nestmates, at both the
individual and colony level, as assumed by the model of
Newey (2011), meaning that errors can be made when the
chemical distance is small. However, these recognition
errors may not be of great significance, because even a high
non-nestmate recognition error rate could be tolerated in a
large population. Based on agent-based models, Axelrod
et al. (2004) predicted that cooperation will be maintained
in groups provided that the non-nestmate discrimination
Z9-alkene distance n-alkane distance
Genetic distance (Fst) Spatial distance (m)
0
20
40
60
80
100
0 200 400 600 800 1000 1200
0
20
40
60
80
100
0 10 20 30 40 50 60 700
20
40
60
80
100
0 2 4 6 8 10 12 14 16 18 20
0
20
40
60
80
100
0 0.1 0.2 0.3 0.4 0.5 0.6
Prop
ortio
n of
Agg
ress
ive
enco
unte
rs (
%)
Fig. 3 The probability of
aggressive encounters in
bioassays was not significantly
correlated (linear regression)
with genetic distance between
colonies (p = 0.154,
R2 = 0.093), spatial distance
(p = 0.997, R2 = 0.000), or
chemical distance between
n-alkane colony profiles
(p = 0.184, R2 = 0.079).
However, the probability of
aggressive encounters was
highly correlated with the
colony Z9-alkene profile
(p = 0.001, R2 = 0.414).
A step function with 95%
confidence belts is shown for
the Z9-alkenes. Tests of
interactions between the
variables did not significantly
improve regressions. Filledpoints indicate polygynous
colonies and open circlesindicate monogynous colonies
Table 2 Multiple regression analysis to compare chemical distance,
genetic distance and spatial distance as predictors of aggressive
response, between conspecific workers of Formica exsecta
Model B SE B b p
Step 1
Constant 39.7 7.85 \0.0001
Z9-alkene distance 1.16 0.23 0.73 \0.0001
Step 2
Constant 27.19 18.71 0.162
Z9-alkene distance 1.09 0.25 0.686 \0.0001
n-alkane distance -1.00 1.03 -0.15 0.345
Genetic distance 66.68 45.99 0.225 0.163
Spatial distance -0.011 0.032 -0.053 0.739
R2 = 0.54 for step 1; DR2 = 0.07 for step 2 (ps = 0.35) Durbin-
Watson 1.314
328 Behav Genet (2012) 42:323–331
123
failure rate (acceptance errors) does not exceed 10%.
However, ant colonies contain hundreds, or thousands, of
individuals, so a non-nestmate will eventually be detected
as it will encounter many ants from that colony. Thus, even
large error rates at an individual level can translate into a
high level of discrimination at the colony level (Johnson
et al. 2011).
It is essential that any mechanism of recognition ensures
a balance between excluding intruders and correctly rec-
ognizing nestmates, because nestmate recognition errors
(rejection errors) would be costly if they led to unnecessary
aggression, exclusion, or death of nestmates. Thus, some
relaxation in the individual chemical distance threshold for
non-acceptance is better than expecting absolute precision,
which would require that all nestmate profiles be consis-
tently identical; a biological implausibility. This study and
previous research has shown that individual F. exsecta
profiles vary little within a colony, with a mean variation of
1–3% for the major Z9-alkenes, which did not correlate
with the level of genetic diversity (Martin et al. 2009). The
total absence of any aggression from resident ants on
mounds towards the control nestmates introductions in all
trials indicates that the F. exsecta nestmate recognition
error rate (rejection error) is 0%. This finding is replicated
in the overwhelming majority of social insect studies
(Johnson et al. 2011). Thus, recognition of nestmates is
always correct, but this can only be guaranteed by using a
threshold for non-recognition that tolerates occasional
acceptance errors since ants like other social organisms
require very strong and good indicators, but small intra-
colony variability exists and this serves as an indicator of
variation without behavioral consequences.
Our study showed that differences in the entire
Z9-alkene profile were highly significant, but differences
between individual Z9-alkenes were not significant. This
supports previous studies (Martin et al. 2008a, b) where
colony specific profiles require several, not single,
Z9-alkenes in order to form numerous unique and distinct
profiles. This finding is important because considerable
debate on recognition system evolution has focused on a
general model of recognition known as ‘Undesirable-
absent/Desirable-present’ or ‘U-absent/D-present’ (Sherman
et al. 1997). The main assumption of such recognition
models is that ‘desirables’ (nestmates) and ‘undesirables’
(non-nestmates) will possess an overlapping set of inde-
pendent recognition cues that are genetically determined
(Getz 1981). However, it is now generally accepted that
such a system would be evolutionarily unstable (Crozier
1986; Rousset and Roze 2007), since any label indicating
altruism would become rapidly fixed in population. The
deciphering of the first ant recognition systems has shown
that intra-specific colony profiles utilize the same suite of
hydrocarbons, but individual components differ in their
relative abundance rather than presence or absence (Martin
et al. 2008a, b; Akino et al. 2004; Greene and Gordon
2007). Despite this the ‘U-absent/D-present’ model is still
employed as a viable approach in recent nestmate recog-
nition studies e.g. (Guerrieri et al. 2009; Couvillon and
Ratnieks 2008). However, a simple presence or absence
system may still operate at the species recognition level,
because thousands of different cuticular hydrocarbons are
produced by ants with no two species sharing the same
profile (Martin and Drijfhout 2009b).
Our study found that the Z9-alkene profile was a sig-
nificant predictor of aggression, but not the n-alkane pro-
file. In this study of 33 colonies the diversity of n-alkane
profiles (0–18%) is significantly smaller than that of
Z9-alkene profiles that range from 2 to 58% (Fig. 3). This
means ants have to be much better at discriminating
between n-alkanes than between Z9-alkenes if n-alkanes
are used as recognition cues. However, the empirical evi-
dence (e.g. Chaline et al. 2005; Dani et al. 2005) indicates
in fact the opposite it true. This shows that comparisons of
total CHC chemical profiles are not valid until the profile
components used in recognition have been identified.
Using a whole profile comparative approach is bound to
lead to unreliable results, especially when a profile contains
numerous minor compounds involved in nestmate recog-
nition, as was found with F. fusca (Martin et al. 2008;
Helantera et al. 2011).
Our study shows that only the distance between entire
Z9-alkene profiles of shared compounds mediates recog-
nition in the ant F. exsecta. The threshold for discrimi-
nating between non-nestmates and nestmates is small, but
not sufficiently narrow that acceptance errors do not occur.
As predicted (Axelrod et al. 2004), the discrimination
system permits occasional errors and allows non-nestmates
to be undetected in individual encounters, but frequent
contacts in highly populous nests means that they are
detected within just a few more encounters (Johnson et al.
2011). Aggression was a nonlinear step function of
chemical distance, with a small change in chemical dis-
tance resulting in a rapid behavioural transition from non-
aggression to overt aggression. The findings of this study
clarify our current understanding of recognition systems,
because they support phenotype matching to a colony
chemical profile, without any major genetic or spatial
components. Thus, our results reflect those typically
observed in social mammals, where differential treatment
is based on rearing associations, i.e. group/colony mem-
bership, rather than genetic relatedness (Holmes 1984;
Holmes 2004). This supports a general interpretation of
social animal organization based on olfactory familiarity
through association (rearing familiarity). However, the
ways these olfactory cues arise and are used may be
fundamentally different, because of the very different
Behav Genet (2012) 42:323–331 329
123
cognitive abilities of mammals and insects. Thus, mam-
mals living in small groups might easily learn the specific
odours of familiar individuals, whereas, this is not possible
in large social insect colonies that can contain thousands of
individuals, so learning the overall colony odour is a far
simpler solution. How that colony odour is established and
maintained is still a subject of ongoing investigation, but
out study shows that any genetic components that might lie
in the chemical profile are obscured by the overall colony
odour. Thus, the establishment of a colony odour in social
insects prevents nepotism and provides an efficient route to
reliable recognition of aliens.
Acknowledgments The authors thank Roger Butlin of Sheffield
University for comments, and Kalle Trontti, University of Helsinki,
funded by Finnish Academy grant 134561 for help in genotyping. The
authors also thank Liselotte Sundstrom of University of Helsinki for
comments and help with the Finnish field work funded by Academy
of Finland grant (206505, 121216) and providing permission to work
on the ant population at the Tvarminne zoological station. This study
was funded by NERC grant NE/F018355/1 and NE/F018088/1.
References
Akino T, Yamamura K, Wakamura S, Yamaoka R (2004) Direct
behavioural evidence for hydrocarbons as nest mate recognition
cues in Formica japonica (Hymenoptera: Formicidae). Appl
Entomol Zool 39:381–387
Axelrod R, Hammond RA, Grafen A (2004) Altruism via kin-
selection strategies that rely on arbitrary tags with which they
co-evolve. Evolution 58:1833–1838
Beecher MD, Beecher IM, Lumpkin S (1981) Parent-offspring
recognition in the bank swallows (Riparia riparia): development
and acoustic basis. Anim Behav 29:95–101
Blomquist GJ, Bagneres A-G (2010) Insect hydrocarbons; biology,
biochemistry and chemical ecology. Cambridge University
Press, New York, p 492
Boomsma JJ, Nielsen J, Sundstrom L, Oldham NJ, Tentschert J,
Petersen HC, Morgan ED (2003) Informational constraints on
optimal sex allocation in ants. Proc Nat Acad Sci 100:8799–8804
Chaline N, Sandoz JC, Martin SJ, Ratnieks FLW, Jones GR (2005)
Learning and discrimination of individual cuticular hydrocar-
bons by honey bees (Apis mellifera). Chem Senses 30:327–333
Chapuisat M (1996) Characterization of microsatellite loci in
Formica lugubris and their variability in other ant species. Mol
Ecol 5:599–601
Couvillon MJ, Ratnieks FLW (2008) Odour transfer in stingless bee
marmelada (Frieseomelitta varia) demonstrates that entrance
guards use an ‘‘undesirable-absent’’ recognition system. Behav
Ecol Sociobiol 62:1099–1105
Crozier RH (1986) Genetic clonal recognition abilities in marine
invertebrates must be maintained by selection for something
else. Evolution 40:1100–1101
Dani FR, Jones GR, Corsi S, Beard R, Pradella D, Turillazi S (2005)
Nest mate recognition cues in the honey bee: differential
importance of cuticular alkanes and alkenes. Chem Senses
30:1–13
Errard C, Hefetz A, Jaisson P (2006) Social discrimination tuning in
ants: template formation and chemical similarity. Behav Ecol
Sociobiol 59:353–363
Getz WM (1981) Genetically based kin recognition systems. J Theor
Biol 92:209–226
Goudet J (1995) FSTAT (vers. 1.2): a computer program to calculate
F-statistics. J Hered 86:485–486
Goudet J (2001) FSTAT, a program to estimate and test gene
diversities and fixation indices (version 2.9.3). (Updated
from Goudet 1995). http://www.unil.ch/izea/softwares/fstat.html.
Accessed 1 March 2003
Greene MJ, Gordon DM (2007) Structural complexity of chemical
recognition cues affects the perception of group membership in
the ants Linephithema humile and Aphaenogaster cockerelli.J Exp Biol 210:897–905
Guerrieri FJ, d’Ettorre P (2008) The mandible opening response:
quantifying aggression elicited by chemical cues in ants. J Exp
Biol 211:1109–1113
Guerrieri FJ, Nehring V, Jørgensen CG, Galizia CG, D’Ettorre P
(2009) Ants recognize foes and not friends. Proc Roy Soc B
276:2461–2468
Gyllenstrand N, Gertsch PJ, Pamilo P (2002) Polymorphic microsat-
ellite DNA markers in the ant Formica exsecta. Mol Ecol Notes
2:67–69
Haag-Liautard C, Vitikainen E, Keller L, Sundstrom L (2009) Fitness
and the level of homozygosity in a social insect. J Evol Biol
22:134–142
Hasegawa E, Imai S (2004) Characterization of microsatellite loci in
red wood ants Formica (s. str.) spp. and the related genus
polyergus. Mol Ecol Notes 4:200–203
Helantera H, Lee Y, Drijfhout FP, Martin SJ (2011) Genetic diversity,
colony chemical phenotype and nestmate recognition in the ant
Formica fusca. Behav Ecol 22:710–716
Holldobler B, Wilson EO (2009) The superorganism: the beauty,
elegance, and strangeness of insect societies. WW Norton &
Company Ltd, London
Holmes WG (1984) Sibling recognition in thirteen-lined ground
squirrels: effects of genetic relatedness, rearing associations, and
olfaction. Behav Ecol Sociobiol 14:225–233
Holmes WG (2004) The early history of Hamiltonian-based research
on kin recognition. Ann Zool Fennici 41:691–711
Johnson BR, van Wilgenburg E, Tsutsui ND (2011) Nestmate
recognition in social insects: overcoming physiological con-
straints with collective decision making. Behav Ecol Sociobiol
65(5):935–944 (online first)
Keller L, Ross KG (1998) Selfish genes: a green beard in the red fire
ant. Nature 394:573–575
Lacy RC, Sherman PW (1983) Kin recognition by phenotype
matching. Am Nat 121:489–512
Martin SJ, Drijfhout FP (2009a) How reliable is the analysis of
complex cuticular hydrocarbon profiles by multi-variate statis-
tical methods? J Chem Ecol 35:375–382
Martin SJ, Drijfhout FP (2009b) A review of ant cuticular hydrocar-
bons. J Chem Ecol 35:1151–1161
Martin SJ, Helantera H, Drijfhout FP (2008a) Colony-specific
hydrocarbons identify nest mates in two species of Formica
ant. J Chem Ecol 34:1072–1080
Martin SJ, Vitikainen E, Helantera H, Drijfhout FP (2008b) Chemical
basis of nest mate recognition in the ant Formica exsecta. Proc
Roy Soc B 275:1271–1278
Martin SJ, Helantera H, Kiss K, Lee YR, Drijfhout FP (2009)
Polygyny reduces rather than increases nest mate discrimination
cue diversity in Formica exsecta ants. Insect Soc 56:375–383
Mateo JM (2009) The causal role of odours in the development of
recognition templates and social preferences. Anim Behav
77:115–121
Morgan ED (2004) Biosynthesis in insects. Royal Society of
Chemistry, Cambridge
330 Behav Genet (2012) 42:323–331
123
Nehring N, Evison SEF, Santorelli LA, d’Ettorre P, Hughes WO
(2011) Kin-informative recognition cues in ants. Proc Roy Soc
Lond B 278:1942–1948
Newey P (2011) Not one odour but two: a new model for nestmate
recognition. J Theor Biol 270:7–12
Ozaki M, Wada-Katsumata A, Fujikawa K, Iwasaki M, Yokohari F
et al (2005) Ant nest mate and non-nest mate discrimination by a
chemosensory sensillium. Science 309:311–315
Palagi E, Dapporto L (2006) Beyond odor discrimination: demon-
strating individual recognition by scent in Lemur catta. Chem
Senses 31:437–443
Queller DC, Goodnight KF (1997) Relatedness Version 5.0. Keck
center for computation biology. Rice University, Houston
Reeve HK (1989) The evolution of conspecific acceptance thresholds.
Am Nat 133:407–435
Roulston TH, Buczkowski G, Silverman J (2003) Nestmate discrim-
ination in ants: effect of bioassay on aggressive behaviour. Insect
Soc 50:151–159
Rousset F, Roze D (2007) Constraints on the origin and maintenance
of genetic kin recognition. Evolution 61:2320–2330
Schmidt AM, d’Ettorre P, Pedersen JS (2010) Low levels of nestmate
discrimination despite high genetic differentiation in the invasive
pharaoh ant. Front Zool 7:20
Sherman PW, Reeve HK, Pfennig DW (1997) Recognition systems.
In: Krebs JR, Davies NB (eds) Behavioural ecology: an
evolutionary apporach. Blackwell Science, Oxford, pp 69–96
Sundstrom L, Keller L, Chapuisat M (2003) Inbreeding and sex-
biased gene flow in the ant Formica exsecta. Evolution
57:1552–1561
Trontti K, Tay WT, Sundstrom L (2003) Characterisation of
polymorphic microsatellite loci for the ant Plagiolepis pygmaea.
Mol Ecol Notes 3:575–577
Vitikainen E (2010) Causes and consequences of inbreeding in the ant
Formica exsecta. Ph.D. thesis, Faculty of Biosciences, Univer-
sity of Helsinki, Finland
Wagner D, Tissot M, Gordon DM (2001) Task-related environment
alters the cuticular hydrocarbon composition of harvester ants.
J Chem Ecol 27:1805–1819
Waldman B (1988) The ecology of kin recognition. Ann Rev Ecol
Syst 19:543–571
Wyatt TD (2003) Pheromones and animal behaviour: communication
by smell andtaste. Cambridge University Press, Cambridge
Behav Genet (2012) 42:323–331 331
123