Journal of Chemical Ecology. Vol. II. No. 8. 1985

MODES OF DEFENSE IN NEMATINE LARVAE Efficiency Against Ants and Birds

JEAN-LUC BOEVÉ and JACQUES M. PASTEELS

Laboratoire de Biologie animale et cellulaire \ Univer.'iité libre de Bruxelles 50, av. F.D. Rooseveit 1050 Bruxelles, Belgium

(Received March 12, 1984; accepted November 29, 1984).

Abstract—Ventral glands are common in nematine larvae (: Symphyla), but they show various degrees of development and are functional for défense only in some species. In those species, volatile irritants are pro­ duced which are effective against ants. Alternative or complementary mech­ anisms against ants are the pubescence of Trichiocampus spp., the foam pil­ . . lars constructed by Stauronema compressicornis, various movements of the abdomen, which occur independently of the glandular sécrétion in several species, immobility of the flat larvae of Nematinus luleus, and burrowing within plant tissues in gallicolous larvae or miners. Glandular development is r not clearly related to the appearancc of the larvae, either cryptic or apose­ matic. The sécrétion, even when it is produced in large amounts by species with well­developed glands, is only moderately efficient against great tits. Bright colors are f'ound in gregarious larvae; thèse were accepted only with reluctance by great tits and sometimes rejected, even species in which the ventral glands are reduced. We suggest that the various volatile irritants se­ creted by ventral glands are aimed primarily against (e.g., ants) and only secondarily against birds.

Key Words—Sawfly larvae, , Hymenoptera, , dé­ fensive sécrétion, ventral glands,mechanical défense, crypsis, aposematism, gregariousness, prédation, ants, birds.

INTRODUCTION

The soft bodied sawfly caterpillars are potentially easy prey for predators and parastoids. Indeed, Benson (1950) gives an impressive list of their potential

1019

0O98-O331/O8OO-lO19$O4..')O/O lO 1985 Plénum Publi.shing Coiporaluin 1020 BOEVÉ AND PASTEELS enemies, among them birds, ants, and parastoids. It is thus not surprising that diverse and elaborate défensive mechanisms are reported in this group (Benson, 1950); however, the évidence for thèse défenses is often anecdotal and detailed studies are few. Chemical défense has been thoroughly studied only in Neodiprion sertifer (Diprionidae) (Eisner et al., 1974), and in Perga affinis (Pergidae) (Morrow et al., 1976). Both species sequester host plant terpenes in, respectively, two and one pouches of the foregut, and they regurgitate thèse compounds when the larvae are disturbed. Recently, a toxic octapeptide was found in the Australian larva Lophyrotoma interrupta (Pergidae) (Williams et al., 1982). Herbivorous vertebrates are frequently poisoned by ingestion of thèse larvae. Nematine (Tenthredinidae) possess medioventral glands which can be everted when the larvae are alarmed. They release a fluid which, in some cases, is odorous (Yussa, 1922; Benson, 1950; Maxwell,1955; Alsop, 1970; Smith, 1970). The composition of the sécrétion has been studied for nine species in the gênera Nematus, Nematinus, Croesus, Pontania, and Pristiphora. The sécrétions contain one or more volatile compounds, such as benzaldehydes, monoterpenes, aliphatic aldéhydes, or acétates (Boevé et al., 1984; Bergstrôm et al., 1984). Thèse or similar compounds are commonly found in défensive sécrétions of insects and were classified by Eisner (1970) as nonspecific toxicants acting as irritants. Unidentified volatile compounds were also detected in Pla- tycampus luridiventris and in additional Nematus, Nematinus and Pristiphora species (unpublished). In this paper we study the interspecific variation in gland development and attempt to correlate it with larval appearance and with the efficiency of défense against two sorts of important predators: ants and birds. For convenience, the term aposematism will be used throughout the text, although this probable func- tion of the bright colors will be documented later in the paper.

METHODS AND MATERIALS

Larvae of 28 species were studied. They were ail collected in the field in Belgium and maintained in the laboratory on their natural host plant. Both sawfly and host-plant species are listed in Table 1. Lorenz and Kraus (1957) and Nigitz (1974) were used to identify the larvae, and Benson (1951, 1952, 1958) for the adults. Morphological studies were conducted on larvae fixed in Dubosq-Brasil and embedded in Paramat. The 7-/xm sections were stained with ferrie trioxy- hematein-phloxin-light green. Gland size was evaluated by measuring the area of the gland, colored with hemalum and mounted in Canada balsam. For each species, one to four last-instar larvae were dissected. The surface measured is in point of fact the surface of the glandular tissue as observed, for example, in DEFENSE IN SAWFLY LARVAE 1021

TABLE 1. APPEARANCE AND FEEDING HABITS OF SPECIES STUDIED"

Appearance Distribution Host plants Investigations

Croesus septentrionalis (L.) A G Alnus, Betula 1,2,3 C. vams (ViU.) C G-S Alnus 1,2,3 australis (Lepel.) C S Alnus, Betula 1,2,3 H. crocea (Geoffr.) A G Alnus, Betula 1,2,3 Mesoneura opaca (Kl.) C S Quercus 1 Nematinus luteus (Pz.) C S Alnus 1, 2, 3 Nematus bipartitus (Lepel.) C s Populus 1, 2,3 A^. hypoxanthus (Fôrst.) C s Populus, Salix 1 N. melanaspis (Hg.) A G Populus, Salix 1,2,3 N. melanocephala (Hg.) A G Salix 1,2,3 N. miliaris (Pz.) A G Salix 1,2,3 A'^ pavidus (Lepel.) A G Salix 1,3 N. salicis (L.) A G Salix 1,2 N. spiraeae (Zadd.) C Ag Aruncus 1,3 N. tibialis Newm. C S Robinia 1,2,3 Pachynematus scutellatus (Hg.) C S Picea 1,3 Platycampus luridiventris (Fall.) c S Alnus 1 Pontania vimimlis (L.) 9 7 Salix 1,3 Pristiphora aquilegiae (Voll.) c Ag Aquilegia 1 P. compressa (Hg.) c S Picea 1.3 P. conjugata (Dahlb.) A 6 Populus l P. erichsonii (Hg.) A G Larix 1 P. palHpes Lepel. C Ag Ribes 1, 2, 3 P. saxesenii (Hg.) C S Picea 1, 3 P. testacea (Jur) A G Betula 1,2,3 Stauronema compressicornis (Fabr.) C S Populus 1,2,3 Trichiocampus ulmi (L.) C S Ulmus 1, 2 T. vimimlis (L.) A G Populus 1,2,3

"A: aposematic, C: cryptic. G: gregarious, S: solitary, Ag: aggregated (see text), 1 : glandular morphol- ogy, 2: tests with ants, 3: test with great tits.

Figures 1A-C, and which represents half the total glandular surface (see Figure ID). This measure was used as a rough approximation of thèse glands' défensive investment. Measurement of the volume of sécrétion would have been impossible in those species with reduced glands, and for which no détectable amount of sécrétion may be collected. Défensive efficiency against ants was observed by placing a larva wifh 20 Myrmica rubra workers in a square container (side: 10 cm). Two minutes were allowed for the ants to discover the larva. During the next 3 min, the number of ants attacking or surrounding the larva was counted every 20 sec on a video- recording of the experiments. This quantifies the interest of the ants for the larvae as prey and is considered to be inversely related to the larvae's défensive 1022 BOEVÉ AND PASTEELS ability. Depending on the availability of the sawfly species, one to six répétitions were made with separate larvae. For some species, additional observations were made with larvae on their host plant. A twig of the plant bearing a larva was then placed in the foraging area of a laboratory nest of Myrmica rubra. To test the efficiency of défense against birds, larvae were ofFered to two caged great tits (Parus major, one maie, one female) collected as adults in the field. In each session, a bird received in succession a mealworm, three full- grown sawfly larvae of the same species, and again a mealworm. Acceptance or rejection was recorded. The limited supply of biological material excludes a précise quantification of the différent larvae's palatability, and thèse tests were only carried out to show up possible trends in the birds' reactions. No bird was ever tested with more than three différent species per day, in order to avoid satiation. No bird ever refused a mealworm. Statistical tests are described in Siegel (1956).

RESULTS

Crypsis and Aposematism. Nematine caterpillars oflFer some of the best examples of both crypsis and aposematism. Two extremely cryptic species are illustrated in Figure 1 (A and B). The larva of Platycampus luridiventris (Figure lA) is green, flat, with latéral expansions, and is tightly appressed to the leaf surface. Thèse larvae are solitary and located on the underside of the leaf, often along the major veins when not feeding. Nematinus luteus larvae (Figure IB) are semicylindrical in shape. They are green with small white dots, mimicking the texture of the leaf The latéral margins are light, counteracting the shadow effect. They are solitary and live on the upper surface of the leaves. Brightly colored larvae are illustrated in Figure 1 (C and D). Larvae of Hemichroa crocea (Figure IC) feed in groups along the leaf-edge. They are light brown with black longitudinal Unes on each side. They usually curl the extremity of their abdomen ventrally. The larvae of Croesus septentrionalis are yellowish with black dots and a black head. They also feed in groups at the edges of the leaves. When disturbed, they raise their abdomens in a typical défensive posture (Figure ID). The appearance, cryptic or aposematic, and the feeding habits, gregarious or solitary, of the difl'erent species studied are given in Table 1. Gregarious habit is significantly correlated with bright coloration and crypsis with solitary habit (x^, P < 0.001). Solitary species, if not extremely cryptic as described above, are at least the same color as the leaf There is no association between type of appearance and taxonomic position; cryptic and aposematic larvae are often found within the same . The larvae of a few solitary species can, however, appear aggregated as a DEFENSE IN SAWFLY LARVAE 1023

Fig. 1. Cryptic (A, B) and aposematic (C, D) larvae of nematines. (A) Platycampus luridiventris, (B) Nematinus luteus, (C) Hemichroa crocea, (D) Croesus septentrionalis. resuit of the distribution of their host plant. Both Nematus spiraeae and Pristi- phora aquilegiae were found on ornamental herbs, respectively Aruncus silves- ter and Aquilegia vulgaris, and a large number of larvae is often observed on a single plant. Nonetheless they feed mostly by themselves and are not brightly 1024 BOEVÉ AND PASTEELS colored. N. spiraeae and its cultivated host plant are introduced North American species. The cryptic larvae of Pristiphora pallipes can form dense populations on gooseberry shrubs in gardens. We do not know how thèse three species are distributed in their natural habitat. The larvae of Croesus varus are found either singly or in small groups depending on their âge. The green full-sized larvae are usually solitary, whereas the younger, also green, larvae are more gregarious. Very young larvae are grouped and blackish, thus contrasting with the substrate. This may be an intermediate between clearly cryptic and clearly aposematic species. The char- acteristics of this species are not fixed but change gradually during ontogeny. The larvae of Pontania spp. live in galls and are thus concealed. Mature galls, however, are sometimes bright red, but the function of this coloration remains unknown. Ventral Glands. Typically seven glands are présent, located midventrally in the first seven abdominal segments. Each gland consists of a pocket in the integument, which at rest is invaginated into the body. The cuticular réservoir is lined by a single layer of glandular cells and is prolonged by a flat and broad duct leading to a slit-shaped opening (Figure 2D). Thèse slits open posteriorly and on those segments with prolegs just behind them. Muscle fibers link both sides and the apex of the gland to the body wall. The gênerai structure of the glands seems quite similar throughout the subfamily, but, as illustrated in Figure 2, considérable variation was encountered in the development of glandular tissue and musculature. Ventral glands are undoubtly used for défense in those species where they are well developed. Disturbed larvae raise and wave their abdomens and evert their glands by blood pressure (Figure 3).Eversion of the glands was observed in Croesus spp., Nematus pavidus, N. melanaspis, N. spiraeae, and in Pristi• phora spp., whereas eversion was never seen in some species with reduced glands, e.g., Hemichroa spp., and Nematus salicis. In those species, we were unable to detect any sign of sécrétion, and their glandular musculature is often reduced to a single pair of fibers (Figure 2B and C). In Figure 4, the species are ranked according to the absolute size of the glandular tissue in full-grown larvae. In species with well developed glands, the first and last glands are smaller than the others. Interspecific comparisons were therefore made by measuring glands from the 2nd to 6th segments. Both cryptic and aposematic species are found among species with or without voluminous glands. No significant différence between the distributions of gland sizes in apose• matic and cryptic species can be demonstrated using the "médian test" (P < 0.23). However, Fisher's exact probability test suggests that crypsis is often associated with smaller glands (< 0.1 mm^) and aposematism with larger glands (>0.1 mm^) (P = 0.046). The size of the glands is not simply proportional to the size of the body, as DEFENSE IN SAWFLY LARVAE 1025

Fig. 2. Ventral glands of three nematine species, same scale; (A) Nematus pavidus. (B) Nematus salicis, (C) Pontania viminalis. (D) Sagittal section through a ventral gland of Nematus pavidus.

demonstrated by devising an index which reduces différences due to larval sizes. The index used was: the glandular surface (in 10"^ mm^) divided by the square of the head capsule width (in mm^). For preserved spécimens, the width of the head capsule is a good measure of the body size: its cubic dimension is strongly correlated with the weight of living larvae (r = 0.91; 243 larvae of varions âges belonging to 12 différent species). The values of the index calculated for penul- timate larvae of Nematus pavidus fall within the range of values calculated for last-instar larvae, which demonstrates that, at least for différent instars of a species, the index éliminâtes variations due to âge and size. In Figure 5, the species are ranked according to relative glandular devel- 1026 BOEVÉ AND PASTEELS

Fig. 3. Sécrétion at the tip ol cveitcd ventral glands (arrow) of Nematus melanaspis. opment. There is no corrélation between glandular development and taxonomic status, as can be seen by the ranks of the différent Nematus species. Further, no significant corrélation can be found between appearance and relative glandular size. Since a corrélation was found between the absolute size of the gland and aposematism (see above), it is the ability to produce large amounts of volatile compounds, independently of body size, which could be functionally linked to bright colors. Défense against ants. The number of ants surrounding the larva increased during the fîrst 2 min after beginning the test and then stabilized. We estimate the efïiciency of défense by calculating the mean number of ants surrounding the larvae after 2 min and at each subséquent count until the end of the exper- iment (see Methods and Materials). This number could in principle be related to the size of the larva. However, there was no corrélation between the weight of the larva and the number of ants around it (r = 0.01 ; 54 larvae belonging to 17 différent species). In Figure 6, the species are ranked according to the mean number of ants surrounding them. Gland sizes are significantly smaller in the most heavily at• tacked species (mean number of ants > 3) than in the less attacked species (mean number of ants < 3) (P < 0.005, médian test). Heavy attacks are as- DEFENSE IN SAWFLY LARVAE 1027

glanduUr surface (lO^^ imn" )

1p 20 30 ^0 50 60 70 80 90

• Pontania viminalis

« Trichiocampus ulmi C

• NemaCus hypoxanthus C

• Nemacus cibialis C

• Nemacus salicis A

• Hemichroa crocea A

• Hemichroa ausCralîs C

• Nemacus melanocephala A

• Platycampus lurîdivenCris C

• Mesoneura opaca C

• Nemacus biparcicus C

• Prisciphora pallLpes C

• Trichiocampus viminalis A

« Nematinus LuCeus C

• Pachynemacus scuteUaCus C

• Stauronema coopressicorais C

9 Prisciphora aquiLegiae C

• Prisciphora conjugaca A

• Prisciphora compressa C

• Nematus miliaris A

• Prisciphora saxesenii C

• Necoacus melanaspis A

# Croesus varus C

• Prise iphora ces tacea A

• Nemacus spiraeae C

• Nemacus pavidus A

# Prisciphora erichsonii A

• Croesus sepCencriona1is A

Fig. 4. Gland size and appearance in nematine larvae. (A) aposematic, (C) cryptic.

sociated with very small glands (<0.05 mm^), whereas few attacks or no at• tacks are associated with voluminous or medium-sized glands (>0.05 mm^) {P < 0.001, Fisher^s exact probability test for two samples). Some species with reduced glands are exceptions to this gênerai rule and are apparently well protected against ants. Also there is no simple corrélation 1028 BOEVÉ AND PASTEELS

Glandular index

5 10 15 20 • • • •

Nemacus saLicis A

Hemichroa auscralii C

Henichroa crocea A

Trichiocanpua ulmi C

Nemacus meLanocephala A

Ncmacus hypoxanchus C

Poncania vininalîs

I Tcichiocampus vininalis A

I Nematus CibiaLis C

• Kesoneuca opaca C

• Nemacus biparcitus C

# Pachynematus scutelLacus C

« Placycatnpus luridivencris C

0 Nenacinus luteus C

^ Nematus mlLiaris A

9 Scauronema compressicomis C

• Prisciphora paLLipes C

• Prisciphora aquilegiae C

• Prisciphora compressa C

« Prisciphora conjugaCa A

• Prisciphora saxesenii C

• Croesus varus C

• NemaCus melanaspis A

• Prisciphora erichsonii A

^ Prisciphora cescacea A

% NemaCus pavidus A

« Croesus sepcentriona1is A

• Nematus spiraeae C

Fig. 5. Relative gland size and appearance in nematine larvae. Glandular index: glan• dular surface (in 10"^ mm^) divided by the square of the head capsule width (in mm^). (A) aposematic, (C) cryptic. DEFENSE IN SAWFLY LARVAE 1029

1 2 3 i 5 5 7 8

) Pristiphora palHpes (6.^)

< Pristiphora cestacea (35.0)

• Scauronema compcessicomis (9.3) '

• Trichiocampus ulmi (2.5)

• Neotatus miLiaris (12.9)

• Nemacus melanaspis (28.7)

• Croesus varus (33.8)

• Trichiocampus viminalis (6.3)

« Nemac inus luteus (6.8)

• Croesus sepcencrionaLis (69.9)

• Hemichroa ausCralis (3.3)

, • Nematus cibialis (2.6)

• Nematus salïcis (2.8)

• NeBiacus melanocephala (3.8)

• Hemichroa crocea (3.0)

• Nemacus bipartitus (4.6)

Fig. 6. Attack by ants (mean number of ants during test) and gland size (in parenthèses, glandular surface inlO~^ mm^) between glandular development and number of ants attacking the larvae {rs = —0.43, NS Spearman rank corrélation coefficient). This suggests that alternative or additional défensive mechanisms exist. Two additional experiments proved the efficiency of the sécrétion: (1) Two larvae of Nematus melanaspis were given in succession to the ants. The glands of the first larva had been emptied the previous day. The mean number of ants surrounding the larva with emptied glands was 10.2, whereas it was only 1.7 for the second larva. (2) Myrmica rubra workers dispersed when presented filter 1030 BOEVÉ AND PASTEELS

papers impregnated with the sécrétion of Croesus septentrionalis, C. varus, Nematinus luteus, Nematus melanaspis, N. pavidus, N. spiraeae, Pristiphora erichsonii, P. pallipes, or P. testacea. The ants always avoid thèse papers and often show signs of excitement or discomfort. Among the alternative défensive mechanisms, the dense pubescence of Trichiocampus larvae seems to deterefficiently the ants. If, however, an ant does succeed in biting a T. viminalis larva, the larva quickly drops from its leaf, thus escaping from further attack. A very différent behavior was observed in Nematus melanaspis, N.miliaris, and Croesus varus. Those larvae grip the leaf firmly with their thoracic legs and are able to dislodge biting ants by movements of the abdomen. Nematus melanocephala raises and moves its abdomen without releasing any sécrétion. In this species défensive behavior is thus purely mechanical and only effective when the ants are few. Nematinus luteus seldom raises its abdomen, but releases a sécrétion smell- ing strongly of citral when disturbed (Boevé et al., 1984). For thèse larvae immobility is an efficient défensive reaction: ants usually step on them without noticing them. If, however, a larva is bitten, movements of the abdomen and émission of the sécrétion repel the ants. A remarkable défensive mechanism is observed in Stauronema compres- sicornis. The larva is well known to build foam pillars (with saliva?) on both sides of the leaf, completely surrounding the (Figure 7). When an ant approaches, the larva moves to the opposite side of the leaf, the ant hits the pillars and immediately retreats rubbing its antennae on the substrate. Larvae separated from their leaf and pillars still seem to deter the ants by producing a sticky sécrétion, the nature and origin of which need further studies. Palatability to Great Tits. Only a few larvae were rejected by the birds, which otherwise accepted the sawfly larvae either as readily as control meal• worms or with various signs of reluctance or discomfort. The great tits eat some larvae only after long delay. Those larvae were tested and rejected several times before being swallowed, or eaten pièce by pièce, whereas the mealworms were always eaten entirely at once. When swallowing a larva, a bird sometimes rubbed its bill on the substrate and/or ruffled its feathers. In ail experiments, the meal- worm given at the end was eaten without hésitation. The results of thèse experiments are given in Table 2. Even though the number of répétitions was small, some clear conclusions can be drawn. On the one hand, the birds readily ate most cryptic larvae in the same way as meal• worms, although some of thèse larvae possess well-developed glands and sécré• tion. Larvae of Croesus varus were the only cryptic ones eaten with some re• luctance, and this will be discussed later. On the other hand, the birds always showed various degrees of reluctance to eat aposematic larvae, even those with extremely reduced ventral glands. The proportions of fuUy palatable larvae are DEFENSE IN SAWFLY LARVAE 1031

Fig. 7. A larva of Stauroncma compressicornis surrounded by spumous pillars.

much higher in cryptic than in aposematic species, and this différence is highly significant {P < 0.001, Fisher's exact probability test for two samples). This indicates that bright colors do indeed have an aposematic function. In most cases, aposematic larvae were, however, eventually swallowed, and we never observed any later sign of discomfort or poisoning in the birds. Neither did we observe, during this expérimentation, any long-lasting avoidance con- ditioning to those larvae. Most aposematic sawfly species were thus only mod- erately distasteful and apparently not toxic to great tits, even though some larvae were eventually rejected. The distributions of gland sizes are not significantly différent in palatable and unpalatable species {P — 0.15, médian test). However, the Fisher test suggests a corrélation between voluminous glands and unpalat- ability: the proportion of species with voluminous glands (>0.2 mm^) is higher in unpalatable species (P = 0.036). The défensive sécrétion from ventral glands 1032 BOEVÉ AND PASTEELS

TABLE 2. PALATABILITY OF NEMATINE LARVAE TO TWO GREAT TITS (I, II)"

Response of the birds Glandular development Appearancc (glandular surface in 10 ^ mm^) I II

Croesus septentrionalis A 86.4 -(b,c) -(a) C. varus C 33.8 -(b,c,d) -(c,d) Hemichroa australis C 33 + + H. crocea A 3.3 -(b,c, d) Nematinus luteus C 6.8 + Nematus bipartitus C 4.6 + + N. melanaspis A 28.7 -(b,d) N. melanocephala A 3.8 -(b,c, d) -(b, c,d) N. miliaris A 15.5 -(d) -(b, c, d, then a) N. pavidus A 40.6 -(b,d) -(b, c, d) N. spiraeae C 36.2 + + N. tibialis C 2.6 + + Pachynematus scutellatus c 7.6 • Pontania viminalis* 1.S • Pristiphora compressa C n£ • P. pallipes c 6.4 + + P. saxesenii c 16.4 + P. testacea A 35.0 -(b,c,d, then a) Stauronema compressicornis C 93 + Trichiocampus viminalis A 6.5 -(b,c,d) -(a)

accepted without reluctance; -: various signs of reluctance shown as specified in parenthèses; a: the larva is pecked and eventually rejected; b: the larva is tested several times and eaten after a long delay, often pièce by pièce; c: the bird rubbed its bill on the substrate; d: the bird ruffled its feathers; * galls were given to the bird and were opened by it; A: aposematic; C: cryptic.

can be effective against great tits but only when it is produced in large quantity. Larvae of Croesus septentrionalis became more palatable and those of C. varus completely palatable, after their sécrétions were first experimentally removed. Both species secrète the pentacyclic monoterpene, dolichodial (Boevé et al., 1984). They are vigorous species with well-developed ventral glands. Interest- ingly, Nematus spiraeae has glands of about the same size as those of C. varus, and also produces dolichodial as a major constituent of the sécrétion (Boevé et al., 1984). A', spiraeae, however, was readily accepted by the birds. We have no straightforward explanation for thèse contradictory results, which could be due to the small number of tests performed. Also, the size of the gland is only a rough approximation of the amount of sécrétion produced. DEFENSE IN SAWFLY LARVAE 1033

DISCUSSION

Défensive inechanisms are very diversifiée! in larval nematines. The ventral gland sécrétion is just one of them, prominent in some species, absent in many others. The occurrence of ventral glands is a common feature to nearly ail ne- matines. Only the non-European species Pikonema alaskensis is devoid of them (Maxwell, 1955). No other function than défense is known for the sécrétion of thèse glands. This chemical défense has thus been secondarily lost in species with reduced, apparently nonfunctional glands. An alternative hypothesis is that small glands could produce minute amounts of very toxic compounds. Up to now there are no data supporting this hypothesis, which seems to us unlikely. Indeed, no sécrétion at ail could be coUected from reduced glands, and we do not know of any insect protected only by trace compounds. Although, as dem- onstrated in this study, it can be effective against some predators, ventral gland sécrétion must have its drawbacks. We do not know which sélective pressures could favor alternative modes of défense. Important factors such as the cost of chemical défense or of the alter• native mechanical défenses, e.g., pubescence of Trichiocampus larvae or the spumous pillars built by Stauronema compressicornis, have not yet been stud- ied. Counteradaptation of predators to the sécrétion could lead to the develop- ment of new modes of défense. Again this aspect of défense remains little stud- ied. It has been reported that some parasitoids are attracted instead of deterred by the défensive movements of the larvae of the sawfly, Neodiprion swainei, which, however, belongs to another family (Diprionidae) and is devoid of ventral glands (Tripp, 1962). Sympatric bright-colored species of similar appearance, some with well- developed glands, others with reduced glands, could be interpreted as a Batesian mimicry complex as suggested earlier (Pasteels, 1976). Our study cannot con- firm or reject this hypothesis. Aposematic species with reduced glands were eaten only with reluctance by the birds. We cannot specify if this is due to some innate or learned response to aposematic colors or if chemical défenses other than the ventral gland sécrétion are présent in thèse species. The constituents of the ventral glands are diverse. They include aliphatic alcohols, aldéhydes, and acétates; benzaldehyde; and varions monoterpenes, cyclic or not (Boevé et al., 1984, Bergstrôm et al., 1984). Thèse are well-known insect défensive compounds, and their efficiency against ants was confirmed by our results. In contrast, thèse sécrétions seem less efficient against birds. Tables 3 and 4 summarize the main results which suggest a higher efficiency against ants. Many species with functional glands were readily accepted by great tits. It is, of course, difficult to draw gênerai conclusions from experiments using only two of the potential predators. We have, however, no reason to suspect that 1034 BOEVÉ AND PASTEELS

TABLE 3. COMPARISONS OF GLAND SIZE DISTRIBUTIONS IN SAWFLY LARVAE: (MéDIAN TEST)

Cryptic versus aposematic NS,P < 0.23 Heavily attacked by ants versus less attacked Higlily significant, P < 0.005 Palatable versus unpalatable to peat tits NS,P ~ 0.15

great tits, of ail the insectivorous birds, or Myrmica rubra, of ail the predacious ants, have developed exceptional reactions towards sawfly larvae. Our results thus suggest that the sécrétion of the ventral glands is aiming primarily at insect predators (e.g., ants) and not at birds. This supports the hypothesis that volatile irritants commonly found in insect sécrétions have evolved against small arthro- pod enemies (Pasteels et al., 1983). Nematine sécrétion needs to be produced in large quantity to show some activity against birds. The proportion of larvae with very large glands is significantly higher in aposematic species than in cryp• tic species. AH brightly colored species were proved to be slightly to highly unpalatable for great tits, which suggests that bright coloration has an aposematic function. Thèse larvae, however, did not appear to be highly toxic, and it has been re- ported that brightly colored larvae were captured in large quan- tities by cuckoos (Coward, 1920), and Croesus septentrionalis larvae by young chaffinches (Benson, 1950). We do not believe that such occasional observations demonstrate that thèse larvae are not protected. Even a slight unpalatability could be sufficient to lower prédation in complex communities, especially when more favorable food is available. Also, if no conditioning of the birds to unpalatable larvae was ob- served during the expérimentation when few larvae were tested at one time, this does not preclude that conditioning to gregarious larvae could occur in nature. Aposematic colors are linked to gregarious habit. This is a common feature in insects (Fisher, 1930). Gregariousness increases the efficiency of the signal and is not compatible with crypsis. Altruistic behavior, i.e., the warning of predators, is more expected to evolve by kin sélection in aggregated siblings issued from a single egg batch (réf. in Pasteels et al., 1983).

TABLE 4. CORRéLATIONS CONFIRMED BY FISHER'S TEST

Aposematism and voluminous glands (>0.1 mm^) /'~ 0.046 Protection against ants and medium-sized or P < 0.001 voluminous glands (>0.05 mm^) Unpalatability to great tits and voluminous F^iO.OSô glands (>0.2 mm^) DEFENSE IN SAWFLY LARVAE 1035

Only two classes of predators were tested against sawfly larvae. Their known enemies are, however, very diverse (Benson, 1950). Among them, parasitoids are of prime importance. Any attempt to study the évolution of dé• fensive stratégies in sawflies should take their rôle into account. Studies on défense against parasitoids are badly needed. Sawfly species with completely différent défensive stratégies are otten tound on the same host plant. For example, the cryptic solitary Stauronema compres- sicornis surrounded by défensive pillars; the gregarious, aposematic, and hairy Trichiocampus viminalis with reduced ventral glands; and the aposematic Ne- matus melanaspis with copious défensive sécrétion, can be found on the same Populus tree. In a complex environment, a given set of sélective pressures does not necessarily lead to the évolution of a single optimal solution, but rather to a finite number of stable equilibrium.

REFERENCES

ALSOP, D.W. 1970. Défensive glands of : Comparative morphology of selected types. PhD thesis, Cornell Univer.sity, 378 pp. BENSON, R.B. 1950. An introduction to the natural history of British sawflies (Hymenoptera: Sym- phyta). Trans. Soc. Bri. Entomol. 10:45-142. BENSON, R.B. 1951. Handbooks for the Identification of British insects; Hymenoptera 2. Symphyta 6, Part 2(a), Royal Entomol. Soc, London. pp. 1-49. BENSON, R.B. 1952. 6, Part 2(b). pp. 51-137. BENSON. R.B. 1958. 6, Part 2(c). pp. 139-252. BERGSTRôM, G., JONS.SON, S., and LANNE, B.S. 1984. Volatile signais in the defence behavior of larch sawfly larvae (Pristiphora erichsonii and P. wesmaeli). XVII Int. Congress of Entomol- ogy, Hamburg Aug. 20-26. Abstract volume 448. BOEVÉ, J.L., BRAEKMAN, J.C, DALOZE, D., HOUART, M, and PASTEELS, J.M. 1984. Défensive sécrétions of Nematinae larvae (Symphyta Tenthredinidae), Experienlia 40:546. CowARD, T A. 1920-1926. The Birds of the British Isles, vols. 1-3. Frederick Warne. London. EiSNER, T. 1970. Chemical défense against prédation in arthropods, pp. 157-217, in E. Sondheimer and J. B. Simeone (eds). Chemical Ecology. Académie Press, New York. EiSNER, T., JoHNESSEE, J.S., CARREL, J., HENDRY, L.B., and MEINWALD, J. 1974. Défensive use by an insect of a plant resin. Science 184:966-999. FiSHER, R.A. 1930. The Genetical Theory of Natural Sélection. Clarendon, Oxford. LORENZ, H., and KRAUS, M. 1957. Die Larvalsystematik der Blattwespen ( und Megalodontoidea). Akademie Verlag, Berlin. MAXWELL, D.E. 1955. The comparative internai larval anatomy of sawflies (Hymenoptera: Sym• phyta). Can. Entomol. (Suppl. I) 87:1-132. MoRROW, P.A., BELLAS, T.E. and EISNER, T. 1976. Eucalyptus oils in the défensive oral discharge of Australian sawfly larvae (Hymenoptera: Pergidae). Oecologia 24:193-206. NiGiTZ, H.P. 1974. Ùberdie Fichten-Nematinen (Hym.. Tenthredinidae) der Steiermark. Z Agnew. Entomol. 75:264-284. PASTEELS, J.M. 1976. Evolutionary aspects in chemical ecology and chemical communication, pp. 281-293. Proc. XVth Int. Cong. Entomol., Washington, D.C. PASTEELS, J.M., GRéGOIRE, J.-C, and ROWELL-RAHIER, M. 1983. The chemical ecology of défense in Arthropods. Annu.Rev. Entomol. 28:263-289. 1036 BOEVÉ AND PASTEELS

I SiEGEL, S. 1956. Nonparametric Statistics for the Behavioral Sciences. McGraw-Hill, New York. SMITH, E.L. 1970. Biosystematics and morphology of Symphyta. II. Biology of gall-making neinatine sawflies in the California région. Ann. Entomol. Soc. Am. 63:36-51. TRIPP, H.A. 1962. The biology of Perilampus hyalinus Say (Hymenoptera: Perilampidae), a pri- mary parasite of Neodiprion swainei Midd. (Hymenoptera: Diprionidae) in Québec, with de• scriptions of the egg and larval stages. Can. Entomol. 94:1250-1270. WILLIAMS, D.H., SANTIKARN, S., OELRICHS, P.B., De ANOELIS, F., MacLEOD, J.K., and SMITH, R. J. 1982. The structure of a toxic octapeptide, containing 4-D-amino-acids, from the larvae of a sawfly, Lophyrotoma inlerrupta. J. Chem. Soc Chem. Commun. 1012:1394-1396. YuASA, H. 1922. A classification of the larvae of the Tenthredinoidea. ///. Biol. Monogr. 7:1-172.