Sexual reproductive success in plants requires male and female development to be highly coordinated (Herrero 2003). Equally, the reproductive fitness of insects that parasitize seeds and seed cones may depend on the coordination of oviposition and plant development (McClure et al. 1998). This is partly dictated by the type of seed plant. The two extant types—angiosperms and gymnosperms—present different opportunities. Pollination and fertilization occur more closely to one another in flowering plants (angiosperms) than in gymnosperms such as conifers. In the latter, the time from pollen arrival to male gamete delivery into the plant egg is a few days in some species, but up to a year in others (Singh 1978). In Douglas fir, it is a gap of six to ten weeks depending on the season (Owens et al. 1991). Female receptivity is delayed while the haploid megagametophyte, which eventually houses the plant eggs, grows and differentiates. Megagametophytes are polarized, with one end producing the gametes and the other end developing a large mass of prothallial cells, the site of future storage reserve accumulation. Megagametophytes are large even before fertilization and it would be surprising if they were not targets for parasites. Although there are many ovules, not every ovule is fertilized, and in some species, such as Douglas fir, seed yield can be very low in some years (Owens et al. 1991). The problem for a seed parasite at the time of oviposition is how to select as many fertilized ovules as possible, as significant storage accumulation only begins in the megagametophyte if fertilization has taken place. Unfertilized ovules are, in theory, undesirable targets because megagametophytes undergo apoptosis a few weeks after plant egg death has occurred.

Parasites are known to influence pollination of flowering plants, thereby improving their chances of ovipositing in or near fertilized seed that will amass nutrient reserves; this is the case of fig wasps (Weiblen 2002). Gymnosperms are evolutionarily more plesiomorphic seed plants than flowering plants; wind-pollination is the rule, with the notable exception of cycads, and with the rare exception of inadvertent, accidental insect pollination. As in the case of the chalcid Megastigmus spermotrophus, insects are left with the choice of ovipositing eggs either in fertilized or unfertilized plant ovules (Hussey 1955). The fact that infestation levels of this species exceeded the expected amount of filled seed led Niwa & Overhulser (1992) and Rappaport et al. (1993) to suggest that the insect must also be choosing unfertilized plant ovules, but no known mechanism could account for this observation. We tested whether female insects were ovipositing in sexually immature (=prefertilization) ovules, in sexually mature but unfertilized ovules, or in successfully fertilized ovules. We studied storage tissue development in both uninfested and infested ovules. We also set out to test whether the interaction between host and parasite was the same. We carried out extensive histological studies on various stages of seed development in uninfested and infested seed collected in France, where this insect is an invasive pest, as well as in its native habitat in western North America to see if the host/insect relationship was similar.

Ovules were processed and stained differently in the two studies. At INRA, ovules were isolated and fixed in FAA (formalin, acetic acid, alcohol). These were processed and embedded in paraffin, cut into 7μm sections, and affixed to gelatin-coated slides. Material was stained with Safranin-Fast Green (general stain). Ovules removed from trees at BCMF were fixed in 2.5% gluteraldehyde in 0.075M phosphate buffer at pH 7.2. The samples were stored at 4°C until required. Samples were processed and embedded in glycol methacrylate (Technovit 7100, Marivac, Canada), then sectioned at 5μm using a Leica SM 2400 sledge microtome with a tungsten carbide knife. Sections were stained with Ponceau Red 2R/Azure Blue for proteins and cell walls, respectively, IKI, or Toluidine Blue O as a general metachromatic stain. All methods followed Gutmann (1995). To verify lipids, material was processed for electron microscopy using previously described methods (Rohr et al. 1989). Thin sections were contrasted with Reynolds lead citrate and viewed with either a Hitachi HU12A or a Philips CM120 transmission electron microscope. Thick sections (1μm) were stained with Toluidine Blue O in 2.5% aqueous sodium carbonate, pH 8.0.

Figure 1 Oviposition of insect eggs into Douglas fir megagametophytes. (a) Female Megastigmus spermotrophus with ovipositor. Bar, 3mm. (b) Insect eggs deposited in central portion of megagametophyte and (c) in a central cell (arrow), the precursor to (more ...)

Oviposition was restricted to sexually immature megagametophytes. Insect eggs were found almost exclusively in megagametophytes in which central cells were differentiating (figure 2—stage 6). Only rarely were insect eggs found in late stage 5 megagametophytes in which cellularization was nearing completion. Cellularization, also known as alveolation, is the stage during which a syncytial megagametophyte develops walls between its hundreds of free nuclei. Although a number of megagametophyte developmental stages apparently overlap with the period in which insects oviposited (figure 2), this variation is due to differences in development between sampled trees.

Figure 2 Developmental stages of Douglas fir megagametophytes against time, emphasizing the period of Megastigmus spermotrophus oviposition. Shading indicates period of insect oviposition. The length of line indicates the range in length of stage. Stages are according (more ...)

The presence of larvae in unfertilized ovules prevented such degeneration. Unpollinated ovules produced mature plant eggs that were not fertilized and which subsequently broke down, but instead of the entire megagametophyte degenerating within a few weeks, the prothallial cells began to accumulate starch, then lipid and protein. The storage reserves available to the larva changed over the course of its development. The megagametophyte's prothallial cells at the time of oviposition were characteristically highly vacuolated and did not have any lipid bodies, starch grains, or protein bodies. After plant egg abortion in unfertilized ovules, the megagametophyte built up starch grains in the central zone. This zone quickly degenerated spontaneously, forming a corrosion cavity, which is the lacuna that normally forms in a megagametophyte. At its margins, multinucleate storage cells developed (figure 3a), in contrast to earlier stages of megagametophyte development, in which mostly uninucleate cells and the occasional binucleate cell were to be found. Cells began to build up reserves in the form of lipid bodies (figure 3b), protein bodies (figure 3c), and starch (not shown) until all the remaining cells of the megagametophyte were no longer vacuolated but filled with storage reserves. Reserve accumulation continued unabated even as the larva continued to grow. Cells closest to the insect cavity, even if not directly consumed, were less full, indicating that reserves were being broken down. Larvae preferred to eat cells from the margins of the cavity (figure 3d), steadily working their way to the edge of the megagametophyte until they filled most of the seed (figure 3e). There was still megagametophyte tissue surrounding the insect at the time of the X-ray, indicating that feeding does not trigger further degeneration. A schematic summary of the differences between parasitized and unparasitized ovules at two different stages is given in figure 4.

Figure 3 Mature storage tissue in parasitized, unpollinated Pseudotsuga menziesii ovules. (a) Eleven week old megagametophyte cells with granular protein accumulation (arrow) and multinucleate cells (asterisk). Bar, 25μm (b) Electron micrograph (more ...)

Figure 4 Types of Douglas fir seed in the four treatments. (a) The schematic developmental scheme shows ovules at stage 6, week 7 in the top row, and those at week 13 in the bottom, when megagametophyte developmental differences were most apparent. (i) Unpollinated, (more ...)

Another aspect of insect influence was found in the pollinated ovules. As larvae growing in the megagametophyte did not prevent fertilization, the resulting plant embryos were eaten by larvae. The loss of these early stage embryos had no effect on megagametophyte development, which, as in the previous treatment, resulted in multinucleate storage cells with abundant reserves of lipid, protein and starch that served as energy stores for the insect larvae.

In unparasitized fertilized ovules, megagametophyte development and embryogenesis resulted in full seed. Lipid bodies and protein bodies began to fill prothallial cells until the entire megagametophyte was composed of cytoplasmically dense, multinucleate cells filled with nutrient reserves. These megagametophytes developed identically to the two previous treatments.

X-rays showed that seed removed from cones could be segregated according to full seed, empty seed, and parasitized seed (figure 4b). In spring, before insect emergence, it was possible to further segregate male or female adult and larvae still in diapause.

These studies carried out in both France and Canada gave identical results. At some stages, phenology varied by a week, but the staged histological investigations of parasitized and unparasitized ovules, that were either pollinated or not pollinated, did not differ in even minor details.

Insect parasitism of Douglas fir megagametophytes alters seed development in two different ways. The first effect is on unpollinated megagametophytes that normally die within a few weeks of the unfertilized plant egg degenerating (Owens et al. 1991), but when parasitized, fail to abort. Instead, they build up storage reserves as if they had been fertilized. The second effect of insect parasites is on ovules that have been fertilized but whose embryos die at an early stage of development because they are eaten by the larvae. Normally, an immediate consequence of embryos dying at an early stage is the breakdown and death of the surrounding megagametophyte (Orr-Ewing 1957). However, after the larva consumes the young embryo, the megagametophyte does not abort, but continues to build up reserves, feeding the larva instead. In both situations, the mechanisms that trees use to avoid investing resources in failing seed have been co-opted by the insect to its own advantage.

Insect–plant interactions are known to involve insect influence of signal transduction in plants, often by manipulation of hormonal signals (Schultz & Appel 2004). How such factors may be influencing programmed cell death in megagametophytes is unknown, as apoptosis in megagametophytes during seed development has only been studied in relation to germination (He & Kermode 2003a,b). During normal Douglas fir megagametophyte development, concentrations of hormones change predictably (Chiwocha & von Aderkas 2002), but the influence of parasitism remains to be determined.

Manipulation of megagametophytes by insects may not be unique to the host–insect pair of Douglas fir and M. spermotrophus. Similar parasitism by species of Megastigmus is probably to be discovered in hosts whose phenology is similar to Douglas fir, such as Larix (Kosinski 1987). In a comparison of published observations of Megastigmus–seed interactions in two of the largest families of conifers, Cupressaceae and Pinaceae, it was found that Megastigmus oviposit in a manner characteristic to each family. Developing megagametophytes were preferred in pinaceaous conifers, but mature megagametophytes in the case of cupressaceaeous ones (Rouault et al. 2004). It is probably that insects may also override abortion in some of these conifers, but this needs to be verified with histological study.

Grissell (1999) categorized a number of Megastigmus species as gall-formers; seed-infesting species were not included, because the morphology and physiology of galls appears, at least on initial inspection, to be different from what occurs in infested seed. Gall inducers cause host-plant tissues to differentiate into novel structures, ranging in complexity from relatively open pits or folds to structures in which the galler is enclosed entirely by plant tissues (Stone & Schönrogge 2003). Recent phylogenetic studies of different groups of gall insects all support galler control for the major aspects of gall morphology (Cook et al. 2002), gall structures representing the extended phenotypes of galler genes (Stone & Cook 1998). Although the overall morphology of insect galls is variable, the inner organization of gall tissues appears quite similar (Shorthouse & Rohfritsch 1992). The gall-former alters the physiological state of plant tissue, particularly that of cells nearest to the feeding larva, which constitute the nutritive tissue, which is maintained in a metabolically active state by the gall-former (Bronner 1992). For example, in virtually all cynipid galls the larval chamber is lined with nutritive cells and bounded externally by a layer of vacuolate parenchyma and a thin layer of sclerenchyma (Stone et al. 2002). Another way to consider gallers is in the context of the nutrition hypothesis, which states that galls provide enhanced nutrition over other feeding modes (Stone & Schönrogge 2003). Using this criterion, is M. spermotrophus perhaps a galler in Douglas-fir seed? The answer is more ambiguous. On the one hand, at least from a histological point of view, the presence of a chalcid larva does not induce a novel structure, because the development of female gametophyte in unfertilized infested seeds is similar to that observed in fertilized ovules without insects. On the other hand, the chalcid larva controls megagametophyte development, preventing the degeneration process, even inducing differentiation of plant storage tissues. A key criterion for answering the question would be to compare proteins found in the gametophyte during normal development with those induced by the gall insect.

This relationship between M. spermotrophus and Douglas fir differs from other tightly linked insect–plant systems. Yucca moths (Pellmyr 2003), senita moths (Fleming & Holland 1998) and agaonid wasps on figs (Weiblen 2002) benefit their hosts by pollinating flowers in which they have oviposited. These plants produce seeds, a certain percentage of which are lost to developing larvae, which are assured of nutrients from the accumulating seed storage reserves. Douglas fir receives no such advantage as oviposition occurs before fertilization. Megastigmus spermotrophus induces seed storage reserve formation, freeing the insect from a dependence on pollination. As a result, such induced ovules only feed developing larvae. Although this is the first report of this type of insect–host relationship in seed plants, it is probably more widespread. It also provides an excellent system in which to study the insect-mediated delay of apoptosis in another organism.

We acknowledge the help of Drs R. Bennett, J. Turgeon, J.-N. Candau, M. El Maâtaoui, B. Anholt and J.-P. Raimbault, as well as Andrea Coulter, Chris Dewdney, Jenny Robb, Stephen O'Leary and Guy Chanteloup. The following agencies provided support for this research: Fonds France–Canada pour la Recherche, Conseil Général de la Région Centre, Canadian Forestry Service, British Columbia Ministry of Forests and the Natural Sciences and Engineering Research Council of Canada.