INTRODUCTION

Formica (Serviformica) japonica Motschoulsky, 1866, is distributed in the Russian Far East, Mongolia, China, Korea, Japan, and the mountains of Taiwan. It is one of the most common ant species in Japan. It builds both monogynous and polygynous colonies that usually consist of a queen(s) and hundreds to thousands of workers and broods. The workers are usually solitary foragers. Their nestmate recognition is based on chemical signals that are believed to be cuticular hydrocarbon (CHC) blends (Yamaoka, 1990). The blends are shared by the colony members but differ among colonies. In the monogynous colony, the queen unifies the CHC blends of individual workers but the mechanism is still unknown (Yamaoka and Kubo, 1990).

In a citrus garden in Kainan City, Wakayama Prefecture, Japan, we sometimes observed that hundreds of F. japonica workers covered the ground and foraged simultaneously. The foraging style appeared different from that of the F. japonica workers we usually observed in Kyoto. The nest of the Wakayama garden colony was wider but more shallow, and contained dozens of inseminated queens. Morphological features of the workers were almost identical to those of F. japonica in Kyoto, but the cuticular hydrocarbon components apparently differed. We therefore suspected that we were observing more than one species.

In general, a chemotaxonomical approach including cuticular hydrocarbon comparison is valuable for identification of sibling or cryptic species (Howard, 1993). The usefulness of the cuticular hydrocarbon comparison as the taxonomical criterion was also confirmed in a recent chemotaxonomical study on the termite genus Reticulitermes in eastern Asia (Takematsu and Yamaoka, 1999). In this study we compared both CHC components and morphological features among F. japonica colonies collected throughout all of the prefecture of Japan (except for Okinawa, where this species is absent), and discuss the possibility that F. japonica includes several sibling species.

MATERIALS AND METHODS

From 1992 to 2001, we collected Formica japonica workers from all Japanese prefectures except Okinawa. At least 10 workers from each of 3 colonies, but many more if possible, were collected to check intracolonial and intercolonial variation of both morphological features and CHC in each district. We also collected F. hayashi workers because it is currently in the same subgenus and is sympatric in distribution with F. japonica (Terayama and Hashimoto, 1996).

Individual live workers were anesthetized by cooling in a refrigerator and immersion in 500 μl of hexane for 5 min. The extract was concentrated and chromatographed on approximately 500 mg of silica gel (230–400 mesh, Merck), and packed in a disposable Pasteur pipette (7 mm dia.). The hydrocarbons were eluted with 3 ml of hexane, then analyzed by gas chromatography (GC) and gas chromatography-mass spectrometry (GC-MS).

For identification of the position of the double bonds in unsaturated hydrocarbons, the separated hydrocarbons were further chromatographed on approximately 500 mg of silica gel, 230–400 mesh, and impregnated with 10% silver nitrate. Then they were successively eluted with 3 ml each of hexane, 1%, 2%, 5%, 10%, 15%, and 30% ether-in-hexane to separate unsaturated hydrocarbons from saturated ones. The unsaturated hydrocarbons were delivertized by dimethyl disulphide (DMDS) to estimate the position of unsaturated C-C bonds (Dunkelblum et al., 1985)

GC analyses were performed on a Hewlett Packard HP6890 GC equipped with a flame ionization detector. An apolar capillary column (HP-1, Hewlett Packard, 15 m length, 0.25 mm dia., and 0.25 μm film thickness) was used for the analyses. Helium was used as a carrier gas, and the column head pressure was 60 kPa. Injection was made directly onto the capillary column through the cool on-column injector at 53°C and the injector temperature was programmed at oven temperature plus 3°C thereafter. The temperature program of column oven temperature was 50°C for 5 min, 50°C to 310°C at 15°C/min, and then held at the final temperature for 5 min.

GC-MS analyses were achieved with an HP6890 gas chromatograph interfaced to a JEOL JMS SX-102A double focusing mass spectrometer at EI mode with 70 eV, and operated with an HP Model 715/64 computer. GC was operated in the same condition as above, but the column head pressure was 18 kPa.

Nei's distance (Ferguson, 1980) was calculated as an index of the similarity of CHC blends for individual workers from a total of 40 colonies. Nei's distance c was defined as c = |a·b|/(a·b) where a and b are vectors. Each vector presents the CHC blend of individual workers that is composed of 86 elements. The elements are the area size for selected FID peaks with amounts greater than 0.1% of the total amount of the CHC's. For a larger value of c, similarity is considered to increase, and vice versa.

A total of 53 hydrocarbon components with relative content that was more than 1% of the total amount in each colony were selected for multivariate analysis. The data were processed by cluster analysis (CA) and principal component analysis (PCA) and were submitted to discriminant analysis (DA) using Mahalanobis distance. All these multivariant analyses were performed with the “Black-Box package” for data analyses in Aoki, 2001.

RESULTS

Identification of CHC components

CHC compounds of F. japonica consisted of alkenes, n-alkanes and various methylalkanes with 23–37 carbons. At least four different types of CHC pattern were confirmed among F. japonica colonies (Fig. 1a–d, Table 1).

Fig. 1

Gas chromatogram of CHC blends of Formica japonica workers collected in Tsuno (Miyazaki Pref.) (a), Kyoto (Kyoto Pref.) (b), Nasu (Gunma Pref.) (c), Hakodate (Hokkaido Pref.) (d), and of Formica hayashi workers (e)

Table 1

Comparison of four CHC patterns in Formica japonica

The CHC components of the workers collected in Tsuno (Miyazaki Prefecture) were with 23–37 carbons. Alkenes contained 7-alkenes and 9-alkenes, and methylalkanes consisted of mono-, di-, and trimethylalkanes (Fig. 1a) (hereafter, named Type 1). Those of the workers collected in Kyoto (Kyoto) were, however, with 23–33 carbons. The alkenes were all 9-alkenes, and methylalkanes were minor components (Fig. 1b) (Type 2). The CHC components of the workers in Nasu (Tochigi) were with 23–35 carbons, of which methylalkanes were mono- and dimethylalkanes (Fig. 1c) (Type 3). Those of the workers in Hakodate (Hokkaido) were also with 23–35 carbons, but the methylalkanes contained trimethylalkanes in addition to mono- and dimethylalkanes (Fig. 1d) (Type 4).

In contrast, the CHC patterns of F. hayashi were all identical among the colonies collected in Hakodate (Hokkaido), Hachinohe (Aomori), Hanamaki (Iwate), Mito (Ibaraki), Suwa (Nagano), Kyoto (Kyoto), Nakamura (Kouchi), Fukuoka (Fukuoka), and Osumi (Kagoshima). The CHC consisted of alkadiene, alkene, and n-alkanes with 25–33 carbons (Fig. 1e).

Intra- and intercolonial variation of the CHC blends

Resemblance of the CHC blends was evaluated as Nei's distance among 10 nestmate workers each from colonies of Miyazaki, Kyoto, Tochigi and Hokkaido Prefectures (all sites had different CHC types). Nei's distance was calculated between all the pairs of the 10 nestmate workers of each colony (45 pairs). Averages±standard error (s.e) were 0.983±0.003, 0.950±0.026, 0.984±0.010, and 0.977±0.015 in the colonies collected in Miyazaki, Kyoto, Tochigi, and Hokkaido, respectively (Table 2). In the same manner, Nei's distance was calculated between all the pairs of workers from different colonies (100 pairs between 2 colonies). The average distance was always smaller between the colonies with different CHC types than between the colonies with the same CHC types, but the latter was smaller than that among nestmate workers (t-test, P<0.001, Table 2).

Table 2

Resemblance of the cuticular hydrocarbon components within and among colonies with different CHC types

Cluster analysis (CA) (Ward's technique) was conducted on CHC data obtained from a total of 80 F.japonica colonies that were pairs of sympatric colonies collected from 40 different localities (Fig. 2). It suggests the existence of four principal clusters (Type 1, 2, 3, and 4) within the 80 F. japonica colonies. Colonies of Type 1 were collected in Miyazaki (Miyazaki), Sakurajima (Kagoshima), Nagasaki (Nagasaki), Kumamoto (Kumamoto), Matsuyama (Ehime), Hakata (Fukuoka), Takeda (Oita), Toyama (Toyama), Okaya (Nagano), and Kainan (Wakayama); those of Type 2 in Shimonoseki (Yamaguchi), Hiroshima (Hiroshima), Kurashiki (Okayama), Souja (Okayama), Takamatsu (Kagawa), Tokushima (Tokushima), Sasayama (Hyogo), Kyoto (Kyoto), Kashiwara (Nara), and Maibara (Shiga); those of Type 3 in Kofu (Yamanashi), Higashimatsuyama (Saitama), Tama (Tokyo), Narita (Chiba), Tsukuba (Ibaraki), Mito (Ibaraki), Fukushima (Fukushima), Sendai (Miyagi), Hachinohe (Aomori), and Morioka (Iwate); and those of Type 4 in Niigata, Murakami (Niigata), Hakodate, Muroran, Rumoi, Wakkanai, Souya, Yakishiri, Nemuro, and Obihiro (Hokkaido).

Fig. 2

Dendrogram from the cluster analyses of the original CHC blends

Principal component analysis (PCA) was also conducted on CHC data obtained from the 80 colonies. The first, second, and third principal components accounted for 36%, 34%, and 24% of the total cumulative variance, respectively. Analysis of these three principal components shows that the colonies can be classified into four groups (Fig. 3). A plot of the first and second principal components, which accounted for 70% of the total cumulative variance, shows that both Type 2 and Type 4 colonies are well separated from both Type 1 and Type 3. In contrast, in a plot of the first and third principal components, Type 1 and Type 2 colonies are well separated from Type 3 and 4 colonies, whereas that of the second and third principal components of Type 3 and Type 4 colonies are well separated from Type 1 and Type 2.

Fig. 3

Scatterplot of the 80 colonies for the 1st-2nd (a) and the 1st-3rd (b) principal components extracted in PCA

As a second step, stepwise discriminant analysis (DA) using Mahalanobis distance was conducted on the CHC data from 80 colonies to determine which variables separate the four groups. The calibration consisted of the group mean matrix and the inversed pooled correlation matrix. Mahalanobis distances between the various groups were calculated for the data sets from which the calibrations were generated. These distances are displayed in Table 3. As a result of DA, a total of 24 hydrocarbon components were chosen as significant discriminators to predict the groups. Regression coefficients and partial F values are shown in Table 3.

Table 3

Regression coefficient and partial F value for each hydrocarbon component as a significant variable for discrimination of four types by Mahalanobis distance

Distribution of colonies with four types of CHC's in Japan

We classified CHC patterns of approximately 400 F. japonica colonies that were collected in Japan into four groups based on DA. Fig. 4 shows the distribution of the colonies with the four different types of CHC. Type 1 colonies (white circles) were mainly distributed in southern Honshu, Shikoku, and Kyushu. Type 2 colonies (black circles) were distributed in southern Honshu, whereas Type 3 colonies (blue circles) are found in central and coastal (Pacific) northern Honshu. Type 4 (red circles) distribution is on the Sea of Japan coast of northern Honshu and Hokkaido. In several prefectures, the four types were sympatric.

Fig. 4

Distribution of four types of CHC blends in Japan

DISCUSSION

Comparison of the CHC in F. japonica revealed that (1) nestmate workers shared almost identical blends that contained all the CHC components (Table 2), and (2) the CHC components were common but the blend ratios differed even among sympatric colonies. These results are consistent with previous studies on the CHC in ants (Howard 1993; Vander Meer and Morel 1998; Yamaoka 1990), i.e., the ant CHC compositions are species specific and their blend ratios are colony specific. In this study, however, we found (3) four distinct types in CHC composition were present in F. japonica specimens collected throughout Japan (Types 1, 2, 3, and 4 in Table 1).

Cluster analyses on the CHC blends of 80 colonies suggested the existence of four principal groups (Fig. 2). PCA reduced the 53 variables to 3 principal components that represented 94% of the total variance. This also allowed us to establish four groups. The 1st, 2nd, and 3rd principal components characterized Type 2, Type 4, and Type 1, respectively. Subsequently, stepwise discrimination analyses using Mahalanobis distance determined 24 CHC components for discrimination of the four groups (Table 3). Classification of F. japonica colonies based on DA indicated that Type 1 was observed in colonies mainly distributed in southern Honshu, Shikoku, and Kyushu, Type 2 in southern Honshu, Type 3 in central and Pacific-coastal northern Honshu, and Type 4 at the Sea of Japan coast of northern Honshu and Hokkaido (Fig. 4). In F. japonica colonies, CHC blends are paralleled by the colonies, separate distributions. It appears that colonies of these four CHC types correspond to geographical pupulations, and that the four pupulations may be sibling species. As shown in Fig. 4, colonies with different CHC types are distributed sympatrically in several area, including Nagano, Niigata, Toyama, and Yamaguchi Prefectures. We have not yet found hybrid types of CHC blends among the CHC types. If the differences of CHC components are due to intraspecific variation, hybrid types of CHC should be found in the places where colonies with different CHC types nest sympatrically. In contrast, if the differences are due to interspecific variation, hybrid CHC types would be seldom found. Further detailed research, especially at boundaries between different CHC types, is necessary to see if such crossbreeding occurs between reproductives of different CHC types.

In ants, there are many cases where workers of different populations are hardly separated by means of ordinary anatomical traits, although reproductive isolation is strongly suggested (Wilson, 1988; Hölldobler and Wilson, 1990). These sibling species are often separated by the differences in the morphological features specific to the reproductive, i.e., queens and males, or by the differences in chromosome numbers, biology, and/or ethology (Crozier, 1977; Crosland et al., 1988; Halliday, 1981; Ward, 1980a,b, 1983; Seifert, 1991). Therefore, the actual number of extant ant species should be much greater than the current number described, because of these sibling species. This is also true for Formica (Vepsäläinen and Pisarski, 1981; Douwes, 1981). Although F. hayashi was formerly placed in the same taxon with F. japonica, slight habitat differences were observed. With this as a start, their morphology was carefully compared, which resulted in the delineation of two species (Kondoh, pers. comm.; Terayama and Hashimoto, 1996).

Although few minor morphological differences were recognized in workers among the four groups, our discussion recognizing distinct groups in F. japonica may nevertheless be justified. Howard (1993) indicated that insect CHC compositions are generally different at species level, and that there is little intraspecific variation. This provides a new dimension of CHC compositions as phenotypes for discrimination and identification of insect species. Such chemotaxonomical approaches have been used with beetles (Golden et al., 1992; Lockey, 1991), moths (Carlson and Milstrey, 1991; Lavine and Carlson, 1991), termites (Kaib et al., 1991; Haverty et al., 1988; Takematsu and Yamaoka, 1999) and parasitic wasps (Espelie et al., 1990). Many of these studies sought a better means to classify and identify sibling or cryptic species, in which morphological differences were few found. In most cases, the CHC indeed appears to be a valuable character for identification, and also a cue for further detailed morphological comparison.

In F. japonica, we propose that the four groups are independent species. Geographical separation among the four different CHC types may be a result of allopatric or parapatric speciation. If so, it is uncertain how it developed. Even if the differences are due to intraspecific variation, however, the phenomenon is still interesting because there are no reports that show such clear variation in the CHC components within species. To test our hypothesis, further studies are necessary f in the following context: 1) comparison of the genetic distance among the four groups using molecular techniques, 2) comparison of the morphological features specific to reproductives, and 3) observation of reproductive barriers between the four groups by ecological and ethological methods.