TERGAL AND STERNAL GLANDS IN ANTS

Chemical signals, or pheromones as they are generally called, play a central role in the complex communication system of ant societies. During the last 20 years a number of exocrine glands have been identified as the anatomical sources for a diversity of pheromones which mediate sexual and social behavior in ants (for reviews see Wilson 1971, Blum 1977, H611dobler 1978). In recent years, however, several hitherto unknown exocrine glandular structures have been discovered in ants and the behavioral functions of some of them have already been determined. In this paper we will review these findings and will report the new results of our comparative morphological study of tergal and sternal glands in ants.


Introduction
Chemical signals, or pheromones as they are generally called, play a central role in the complex communication system of ant societies.
During the last 20 years a number of exocrine glands have been identified as the anatomical sources for a diversity of pheromones which mediate sexual and social behavior in ants (for reviews see Wilson 1971, Blum 1977, Holldobler 1978).
In recent years, however, several hitherto unknown exocrine glandular structures have been discovered in ants and the behavioral functions of some of them have already been determined.In this paper we will review these findings and will report the new results of our comparative morphological study of tergal and sternal glands in ants.

Material and Methods
For histological investigations live specimens were fixed in alco- holic Bouin (Dubosqu Brasil) or Carnoy (Romeis 1948), embedded in methyl methacrylate, and sectioned 8/x thick with a Jung Tetrander I microtome (Rathmayer 1962).The staining was Azan (Heidenhain).The SEM pictures were taken with an AMR 1000 A Scanning Electron Microscope.For some of the species which could only be identified to the generic level, the respective number is given of the voucher specimens, which are deposited in the ant collection of the MCZ (Harvard University).
*Manuscript received by the editor May 3, 1979.

Results
Tergal glands a. Pygidial gland In a detailed anatomical study of Myrmica rubra Janet (1898) described a pair of clusters of a few glandular cells, located under the third gastric tergum.Each cell sends a duct through the intersegmental membrane between the third and fourth gastral terga.We discovered a similar, but considerably larger paired glandular complex at the same anatomical position in Novomessor cockerelli and N. albisetosus (Holldobler et al 1976).Kugler (1978)   recently investigated a number of myrmicine ants and in many of them he found the gland, which had "distinct reservoirs, produced by invagination of the intersegmental membrane between abdominal tergum 6 and tergum 7 (pygidium)".Kugler suggested that these glandular organs be called pygidial glands.We accept this terminology, because it describes the anatomical designation of the organ more precisely than the term "dorsal gland" or "tergal gland", originally suggested (Holldobler et al 1976, Holldobler and Haskins  1977).However, it has to be pointed out that the pygidium of the ants (the last exposed tergum) is the 7"" abdominal tergum and is not homologous to the pygidium of the Coleoptera (S"' abdominal tergum).Hence, the pygidial glands of ants are not homologous to the pygidial glands of Coleoptera.
In a previous study (Holldobler and Haskins 1977) we found pygidial glands with large reservoirs in several ponerine and myrmeciine ants {Amblyopone, Paraponera, Ectatomma, Odontomachus, Pachycondyla, Platythyrea, Rhytidoponera, Myrmecia) and we demonstrated that the virgin females oi Rhytidoponera metallica attract males by the release of a pheromone from these glands.In his anatomical studies of Rhytidoponera metallica and R. convexa, Whelden (1957,1960) described a pair of cell clusters each comprising 8-15 glandular cells.Each cell sends a duct through the membrane connecting the 6"' and 7"" abdominal segments.We are now certain that Whelden already had discovered the pygidial gland in Rhytidoponera; his histological methods, however, may not have enabled him to detect the large reservoirs associated with the glandular cell clusters.Similar paired glandular structures were found by Whelden (1957) in the ponerine species Stigmatomma {=Amblyopone) pallipes.
The new anatomical investigations presented in this paper reveal that the pygidial glands are much more common in ants than previously assumed.Usually the organ consists of a pair of lateral clusters of glandular cells, each cell sending a duct through the intersegmental membrane between the 6'*' and 7"" abdominal terga.Depending on the species, the intersegmental membrane can be invaginated to different degrees, so that it can form a more or less voluminous reservoir (Fig. 1, 2, 3, 4).If no reservoir is present, the glandular structures can easily be missed during the dissection and histological sectionings are therefore required to determine whether or not the pygidial gland is present.As we have already indicated for Novomessor and as confirmed by Kugler (1978) for several other myrmicine species, the pygidial gland can be associated with a special cuticular structure on the pygidium (7"" tergum), (Fig. 5,6,7).Our histological studies demonstrated, however, that the absence of such structures does not necessarily indicate the absence of Schematic illustration of glandular cells that send ducts through the intersegmental membrane.When the membrane is increasingly more invaginated, it forms an increasingly larger reservoir (a to c).

Psyche
[December pygidial glands.Thus, in several myrmicine species, in which we (Holldobler et al 1976) and Kugler (1978) previously assumed the pygidial gland to be absent, we find that we now detect this organ by histological methods.Tables la and lb list the species of the major subfamilies that we investigated histologically and indicate the type of tergal glands found.
b. Postpygidial gland Dorsal glandular structures which open posteriorly to the py- gidial gland, between the 7"" and 8"' abdominal terga (spiracular plate), we call postpygidial glands (Fig. 8).Whelden (1957,1960) described glands in the 5'*" gastral segment of Rhytidoponem convexa and R. metallica.For R. convexa he writes: "Even the largest of these is less than half the size of the fourth-segment glands ....In the extreme case, there may be but a single gland cell on each side.It is often difficult to distinguish such a unicellular gland from an oenocyte, despite the usually distinct difference in size of the two cell types.Only the identification of a duct certainly distinguishes such a gland from the ductless oenocyte.In many individuals this second pair of glands could not be found".Whelden (1960) makes similar statements for R. metallica.Our results are somewhat different.We found well developed postpygidial glands in the 4 species of Rhytidoponera investigated (Table la).In all specimens we found a pair of clusters of glandular cells.Each cluster contains about 15-20 cells and each cell sends a duct through the interseg- mental membrane close to the spiracular plate (Fig. 9, 10).In some ant species the postpygidial gland consists of only a few glandular cells, in others the postpygidial gland is associated with a well developed reservoir (Table la, lb).

Sternal glands
In several species we discovered intersegmental sternal glands (Table 2).They can consist of a few glandular cells that send their channels through the intersegmental membrane, or of large clusters of glandular cells associated with voluminous reservoirs.These reservoirs are formed by invaginations of the intersegmental membranes (Fig. 8).

Other abdominal glands
The glandular venom apparatus of ants is composed of the Dufour's gland (alkine or accessory gland) and the poison gland.
Although the venom apparatus of ants is very well studied (see reviews by Maschwitz and Kloft 1971, Blum and Hermann 1978 a,   b), other glands, such as Koschevnikov's gland (sting gland), Bordas's gland and sting sheath glands, known from other Hymenoptera, have not been firmly established in ants.Koschevnikov (1899) found in honeybees and Vespa paired clusters of glandular cells located laterally near the intersegmental membrane between the quadrate plate and the spiracular plate.
Each individual cell sends channels into gathering ducts, which connect with the intersegmental membrane.Altenkirch (1962) found similar glands in most Apidae that she had studied.
There are indications that this gland might also be present in some species of the primitive ant subfamilies Myrmeciinae and Ponerinae.Whelden (1957) described a pair of clusters of gland cells, located slightly dorsally on each side of the sting of Stigma- tomma {=Amblyopone) pallipes.Each cell sends a "rather tortuous duct . . .down and inward, to open through a membrane which is above the sting".Robertson (1968) found "sting glands" in Rhytidoponera "toward the region of the triangular plate, where they are attached to the intersegmental membrane".She described similar glands in Bothroponera sp.{=Pachycondyla), Leptogenys sjostedji and Myrmecia gulosa.In the latter species the glands are described as "two well formed masses of gland cells, each cell attached to the intersegmental membrane in the region of the triangular plate by a long, simple, cuticular duct".  A) lists the species in which we found paired glandular structures, closely resembling the "sting glands" described by Robertson.In all cases the glandular cells are located near the triangular plate, and from each cell a rather long duct leads downwards and opens through a membrane near the base of the sting (Fig. 15).Although we could not precisely locate the openings of the ducts, we assume they opened in the sting chamber.
Psyche [December   Altenkirch (1962) and Maschwitz (1964) discovered a so-called sting sheath gland in several bee species.It consists of a palisade epithelium located in the sheath valves.In some ant species we found a distinct palisade epithelium in the sheath valves and/ or single glandular cells with long individual ducts (Table 3 (B), Fig.   16, 17).Janet (1898) describes similar single gland cells, located near the sheath valves, in Myrmica rubra.
Bordas's glands, as they were described by Bordas (1895) in Terebrantia and reexamined by Rathmayer (1962) in several sphecid wasps, could not be identified in ants, although some of clusters of the single gland cells which send their ducts through the membrane of the sheath valves could be related to the Bordas's glands.It is obvious to us that the glandular structures, associated with the sting apparatus of ants, need to be investigated in greater detail in future studies.
In several ant species (Table 3 (C)) we found a highly developed palisade epithelium in the 7"' sternum (Fig. 8).It is especially conspicuous in several Leptogenys species and in the army ants Eciton and Neivamyrmex, but it is not strongly developed in Dorylus.In the dolichoderine species and in Aneuretus this epithe- lium seems to be closely associated with the sternal gland (Pavan's gland).
In the African weaver ant (Oecophylla longinoda) we discovered a sternal gland under the 7*"' sternum, which is quite different from the glandular epithelium described above (Holldobler and Wilson  1976, 1978).This structure consists of an array of single glandular cells that send short channels into cuticular cups on the outer surface of the sternite.In none of the other formicine species investigated, listed in table lb, did we find this type of sternal gland.But in Camponotus sericeus we detected different clusters of glandular cells in the last sternum.Each cell sends a long channel through the intersegmental membrane near the vagina into the ventral part of the "sting chamber".We discovered similar paired glandular cell clusters in most myrmicine species we investigated.The gland is especially distinct in Novomessor and Veromessor, where the glandular cell channels penetrate the membrane near the vagina (Fig. 18).
The function of the intersegmental glands: The functions of most of the glandular structures described in this paper are not yet known, but in a few species the function of the pygidial gland has already been identified.In Novomessor cockerelli and N. albisetosus the strongly smelling secretion of the pygidial glands releases a "panic alarm" response in workers, apparently specifically designed against army ant predation (Holldobler in prep.).Kugler (in press) demonstrated that in Pheidole biconstricta the pygidial glands produce an alarm-defense secretion.A quite different function has been discovered in Rhytidoponera metallica.
Here the wingless virgin females attract males by the release of a pheromone from the pygidial gland (Holldobler and Haskins 1977).
Since Rhytidoponera workers also have a well-developed pygidial gland and are attracted to its secretions, we believe we have not yet discovered the whole functional spectrum of this organ.In Leptogenys chinensis, Maschwitz and Schonegge (1977) demonstrated that the pygidial gland secretions serve together with poison gland substances as a recruitment trail pheromone.We obtained similar results when we recently reexamined the anatomical source of the trail pheromone of Pachycondyla {=Termitopone) laevigata.This ant species conducts well organized predatory raids on termites.
During raiding the workers move in a single file, one closely behind another, along a powerful trail pheromone laid down by leading scout ants.Blum (1966) has identified the hindgut as the source of this recruitment trail pheromone.We cannot confirm his findings.
In our experiments with artificial trails laid with extracts from several abdominal glands, only the pygidial gland secretions released massive trail-following behavior in P. laevigata (Holldobler and Traniello in prep.).A careful observational study of the traillaying behavior of P. laevigata workers revealed that not the anus but rather the pygidial gland is dragged over the ground.Although the pygidial gland of P. laevigata has no definite reservoir, it is very well developed and is associated with an elaborate cuticular structure on the 7"' tergum (Fig. 19,20).The glandular secretion is apparently stored in the many cavities of this structure.When trailing, the ants rub this structure with its special applicator surface over the ground and deposit thereby the trail pheromone.Traniello (pers.communication) observed species of Odontomachus during nest emigrations performing the same trail laying behavior.We suspect that in Odontomachus also the pygidial gland secrets a trail pheromone (Fig. 21).
In Bothroponera {=Pachycondyla) tesserinoda we previously analyzed the signals involved in the tandem running recruitment technique (Holldobler et al 1973, Maschwitz et al 1974).We Psyche [December discovered that the cues responsible for "binding" the follower behind the leader ant include both a surface pheromone and mechanical stimuli.Although we could extract this surface phero- mone, we were not able to identify its anatomical source; all experiments with secretions from the known exocrine glands had negative results.After the recent discovery of the pygidial gland in Pachycondyla we have begun to conduct tandem running experi- ments with Pachycondyla crassa* and P. harpax*, using dummies contaminated with pygidial gland secretions.Our preliminary re- sults strongly indicate that pygidial gland substance might be the source of the tandem running pheromone in these species.
In the doryline army ants raiding and emigrations are conducted along chemical trails deposited by workers.For Neivamyrmex, Watkins (1964) and Watkins et al (1967), and for Eciton hamatum, Blum and Portocarrero (1964), identified the hindgut as the source of the trail pheromone.In addition, Chadab and Rettenmeyer   (1975) and Topoff and Mirenda (1975) demonstrated that besides the relatively long-lasting hindgut trail-substance, other signals (possibly more volatile secretions) are involved in the organization of "mass recruitment" in Eciton and Neivamyrmex.
We believe that our morphological investigations provide new possibilities for the analysis of chemical communication in army ants.Both Neivamyrmex and Eciton have large pygidial glands with distinct reservoirs (Fig. 22, 23).The postpygidial gland is smaller, but still considerably larger than in most of the other investigated speciesf.In both army ant species the 7"^t ergum is relatively small.
Therefore, the reservoirs of the pygidial gland and postpygidial gland open directly above the anus at the abdominal tip (Fig. 23).In workers (all castes) of Eciton the dorsal membrane near the exits of the reservoir of the pygidial glands is conspicuously modified to a brush-like structure (Fig. 24).These morphological features strongly * P. crassa was observed tandem running by W. L. Brown, Jr. (pers.communication) at the western base of Ubombo Mts., Zululand, and by B. Holldobler in Shimba Hills Reserve (Kenya), P. harpax was observed tandem running by S. Levings (pers.communication) on Barro Colorado Island, Panama.
t Whelden (1963) described two glands at the extreme posterior end of the gaster of Eciton burchelli workers.Although the description is not very accurate, from his drawings we can conclude that he found the pygidial gland and postpygidial glands.
suggest that the tergal glands might be involved in the chemical trail communication of army ants.We have begun to test this hypothesis with Eciton hamatum.The pygidial gland secretion of E. hamatum has a strong, characteristic smell.The secretion is probably skatole (Traniello pers.communication), the substance that gives army ants their typical "fecal odor".Recently Brown et al (1978) demonstrated that skatole is an effective growth inhibitor for bacteria and fungi and repels insectivorous snakes (Watkins et al 1969).Our first, preliminary tests demonstrated that Eciton workers follow artificial trails drawn with crushed pygidial glands.When we simultaneously offered trails drawn with hindgut contents and pygidial gland secretions, the latter were significantly preferred during the first minute.When we used trails drawn with secretions of the poison gland or Dufour's gland as controls, the ants always followed the pygidial gland trails.We have to stress, however, that these experiments must be considered pilot tests.The preliminary results, however, are striking enough to warrant a more detailed investiga- tion in the future.It is interesting to note that the anatomy of the pygidial gland in the African army ant, Dorylus molesta, is quite different from that of Eciton and Neivamyrmex (Fig. 25).In this species the T^tergum is considerably larger than in species of the latter genera, and the reservoirs of the pygidial glands do not open at the abdominal tip.In Dorylus, however, we found single glandular cells with channels opening directly at the anus, a feature we have not detected in other ant species (Fig. 25, 26).
Finally, our morphological study of the pygidial gland of Ve- romessor pergandei has led to results that are suggestive of the function of this organ.In this species the 7"" tergum is relatively small, and as a result the large reservoirs of the pygidial glands open at the tip of the gaster (Fig. 27).Veromessor forages in well- organized columns (Went et al 1972; Wheeler and Rissing 1975;   Bernstein 1975).Several observations suggest that these foraging columns are organized by a trail pheromone, though no trail pheromone gland has yet been identified.Clearly, the large pygidial gland has to be considered as a possible source for the trail pheromone.
The function of most of the newly discovered sternal glands is unknown.Only in Paltothyreus tarsatus could we demonstrate experimentally that foragers lay a recruitment trail with sternal gland secretions (Holldobler in prep.).

Conclusions
Since we first found the pygidial gland widespread in the sub- famihes Myrmeciinae and Ponerinae, we speculated that this gland might be a primitive monophylogenetic trait in ants generally (Holldobler and Haskins 1977).The results reported in the present paper fully confirm this assumption.A well-developed pygidial gland was found in the most primitive ant, Nothomyrmecia macrops, and in representatives of all major subfamilies except in the Formicinae (Table lb).We agree with Kugler (1978) that the "anal glands" of the Dolichoderinae and Aneuretinae are homologous to the pygidial glands of other ant subfamilies.Considering the variation in the morphology of the pygidial glands, even within a single subfamily, we think that the morphological variation of the "anal glands" of dolichoderine and aneuretine species does not warrant a separate terminology.In fact the term "anal gland" is misleading, because the glands do not exit from the anal opening of the gaster, as is sometimes inferred, but between the 6"' and 7"" abdominal terga (Fig. 28).This was clearly demonstrated by Pavan and Ronchetti (1955).It is our view and also Kugler's (pers.communication) that the "anal glands" should be called pygidial glands.Kugler (1978) concluded from his comparative studies of myrmi- cine species that usually those species that have reduced or modified stings also have well-developed pygidial glands.He assumes that the pygidial gland replaces the sting apparatus as a chemical defense device.Our finding that well-developed pygidial glands occur in Pogonomyrmex, a genus with a very effective sting apparatus, and in many stinging ponerine species, does not support Kugler's conclusions.

Table la
List of species of the poneroid complex (Taylor 1978) that were investigated histo- logically, and the types of tergal glands found.When the histological series was incomplete and we could not make a definite statement, the column is marked with "?".When the cuticular structure on the pygidium was only slightly sculptured, we marked the column with "-(+)"• Subfamily/ Species     Above: Sagital section through the pygidial gland of Leptogenys pavesii (GC = glandular cells; CH = channels of GC; CS = cuticular structure on the f' tergum).Below: Close-up view of sagital section through cuticular structure of 7' tergum of Leptogenys neutralis.Above: Sagital section through pygidial gland of Leptogenys regis.
Below: SEM photograph of cuticular structure of 7'" tergum of L. regis.The tergura is cut so that the well-structured cavities under the surface can be seen. (Lettering as in Fig. 5).(Lettering as in Fig. 5; sg = sternal gland; ge = glandular epithelium; ppg = post- pygidial gland; pg = pygidial gland).Above; Sagital section through pygidial gland and postpygidial gland, with large reservoirs of Rhytidoponera perthensis.Below: Postpygidial gland only; note the glandular cells with channels. (Lettering as in Fig. 8; R = reservoir).
(Lettering as in Fig. 5; R = reservoir).SEM photograph of the sternal gland area between the 6'" and 7'" sterna of Paltothyreus tarsatus.The picture below is an enlargement of the lighter rectangle in the picture above.The openings of the glandular cell channels in the intersegmental membrane are clearly visible.
Above; SEM photograph of the glandular cell openings of the sternal gland of Paltolhyreus larsatus.Note the filament-like protrusions of the interseg- mental membrane.Below: Sagital section through the same area. (Lettering as in Fig. 5).
(Lettering as in Figs. 5 and 8; R = reservoir).Above: SEM photograph of tip of the gaster of Ecitori hamatum.
Eciton hamatum: SEM photograph of the intersegmental membrane between 6"' and 7"' terga, where the pygidial glands open.Below: Close up of the brush- like structure.Above: SEM photograph of abdominal tip of Liometopum apiculatum."PG" indicates where the pygidial gland opens between 6"^a nd T"" terga.

Figure 7 .
Figure 7.SEM photograph of cuticular structure of 7'" tergum of Plectroctena strigosa.Above: View of the anterior part of the tergum.The cuticular grooves are clearly visible.Below: Close-up view of cuticular structure; the tergum is cut, so that the cavities beneath the grooves are visible.

Figure 11 .
Figure 11.Above: Sagital section through the pygidial gland of Platythyrea cribinoda.The large reservoir is folded in this section.Each cell cluster contains 50-70 cells.Below; Section through postpygidial gland of P. cribinoda: each cluster contains 15-20 cells.(Lettering as in Fig.3).
Figure 19.SEM photograph of the 7'" tergum of Pachycondyla (=Termitopone) Glandular applicator surface (GA) is shown in greater detail in the picture laevigata below

Figure 21 .
Figure 21.Sagital section through the gaster of Odontomachus haematotJa,showing the well developed reservoir of the pygidial gland (PG).

Table lb
List of species of the formicoid complex (Taylor 1978) that we investigated histo- logically, and the types of tergal glands found.

Table 2
List of species of tne poneroid complex in which we found intersegmental sternal glands.CH = channels of glandular cells.Below: Pygidial gland of Myrmica americana: No distinct reservoir, 10-15 cells in each cluster.