We collected adult Oxidus gracilis millipedes from a corn field at the Russell E. Larson Agricultural Research Center at Rock Springs, in Centre County Pennsylvania, USA, during summer and and fall of 2016.

After summer storms, moist soil and high humidity provided for easy surface collection of adult Oxidus gracilis. We collected approximately 100 adults on 28 July 2016; once in the lab, we kept the millipedes in a clear plastic container with a 1-2 cm layer of field soil and a handful of moistened straw. We maintained the colony of 100 adults for four weeks; after three weeks (18 August 2016) we observed clusters of eggs and second instar larvae. However, it was difficult to maintain the colony under proper humidity - too dry and the colony desiccated, too wet and fungus overgrew the colony. The majority of the colony was lost by mid-September; interestingly, this coincided when adult field populations began to decline. In their hypothesized native range of Japan, Oxidus gracilis has been observed to have a similar life history - millipedes mature in summer, lay eggs in fall, and overwinter as late larval stages (Murakami 1962, Murakami 1966). It is probable that their annual life cycle means field-collected adults will die in the fall, regardless of rearing conditions. In early October, although populations were much sparser, we collected additional adults and juveniles for microscopy and to re-start the lab colony.

Adult specimens collected in September were fixed in 80% ethanol. Upon dehydration in ethanol, the cyanide gland became impossible for us to remove – likely due to the gland tissue collapsing. Because of this observation, adult specimens collected in October were kept alive until dissection. At the time of dissection, we placed the millipedes in a petri dish containing 0.1 M dibasic phosphate buffer (ph=7.4) and quickly severed the head from the body using dissecting scissors. Timing of this process seems to be critical; if the millipede is disturbed too much before the dissection, the glands’ storage chambers may be emptied which makes it hard to see the clear gland tissue against the fat body.

After decapitation, the body rings were carefully separated from each other and from the digestive system. Only the segments containing cyanide glands (rings 5, 7, 9, 10, 12, 13, and 15-19) were dissected further under light microscopy; before each ring dissection we checked for the gland ozopore on each keel to ensure we were dissecting rings with cyanide glands (Suppl. material 1).

For each ring dissection, the dorsal tergite and ventral sternite were bisected so each keel could be dissected independently. Care was taken to remove fat body and cuticle to reveal the storage and reaction chambers of each gland.

We use two, fifth stadium juvenile millipedes for additional imaging. Noticing how transparent the cuticle was, and knowing how fluorescent the storage chambers were from the adult dissections, we did not attempt to fully dissect out the glands. From one juvenile we isolated body rings with ozopores and we left the other juvenile largely intact, only severing the head to reduce movement.

Specimens were examined with an Olympus FV10i Confocal Laser Scanning Microscope using two excitation wavelengths: 473 nm, and 559 nm. We detected autofluorescence using two channels with emission ranges of 490–590 nm (green pseudocolor), and 570–670 nm (red pseudocolor), respectively. We generated volume rendered micrographs and media files with ImageJ (Schneider et al. 2012) using maximum intensity projection.

Some glands were easily identified under a light microscope by a characteristic brownish-yellow droplet suspended within the storage chamber. Droplets like this have been observed across polydesmids and are presumed to be the precursor mandelonitrile (Duffey 1981); interestingly, it is still unknown how exactly these droplets are stablized. The quick release required for successful defense limits the likelihood that the percursor is stored as either a glycoside or emulsion (Duffey 1981); but without stabilization, the madelonitrile could react and internally generate cyanide and cause damage to gland tissue. Possibly, the hydrophobicity of mandelonitrile is enough to stabilize the droplet in an aqueous environment, or other compounds (benzoic acid, fatty acids, and fatty acid esters) may play a role in stabilization.

Most storage chambers were difficult to see or not found at all, likely due to destruction of the delicate storage chambers during dissection, or the millipedes evacuating their storage chambers upon disturbance.

We successfully dissected out a storage chamber from one cyanide gland which we imaged on the CLSM using autofluorescense (Figs 1, 2). As expected, the millipede exoskeleton brightly fluoresced and we were able to observe the thin wall of the storage chamber. We were initially surprised to see how brightly the gland extract fluoresced because it is thought to be a mixture of only of small molecules, while arthropod autofluorecense is usually attributed to large, complex molecules (Haug et al. 2011). It is possible that the strong autofluorescence of the gland extract is based on the aromaticity and cyanogenic structure of mandelonitrile and its derivatives (Duffey 1981).

We were unable to observe the cyanide glands from the dissected juvenile millipede; it is possible the gland structures are even more delicate in juveniles, and therefore more prone to tearing during dissection. However, because the cuticle is transparent in the 5^th larval stadium, we successfully imaged the strongly fluorescing storage chambers through the cuticle of the intact juvenile millipede (Figs 3, 4). In one case, we were also able to observe the reaction chamber and valve (Fig. 4).

Upon dissection, we observed some of the adult specimens contained numerous nematodes (Suppl. material 1). O. gracilis can succumb to nematode infection, but have been shown to survive low level infections of an unidentified rhabditid nematode (Poinar and Thomas 1985).

Of course, agricultural systems aren't pristine – they've been invaded by both destructive species (slugs and other crop pests) and beneficial species (carabid beetles, earthworms). We are interested in how O. gracilis fits into this novel food web - how do they interact with pests, predators, and other decomposers? Some farmers worry that the millipedes may be damaging their crops, although O. gracilis refused to eat live plants in laboratory experiments (Causey 1943).

We chose to image the cyanide glands of Oxidus gracilis because the defenses of this invasive species potentially tie into their interactions with generalist predators found in central Pennsylvania. Around 90% of plant biomass is recycled through decomposers, meaning the bulk of plant-based energy travels to predators not through herbivores, but through decomposers (Allison 2006, Polis and Strong 1996). More and more, ecologists recognize the importance of how ‘brown food webs’ link to the more frequently studied 'green food webs' (Wolkovich et al. 2014). In an agricultural context, it is important to understand how generalist predators interact with decomposers; if generalist predators can depend on decomposers as alternative prey, then the decomposer community can play an indirect role in controlling pest populations (Symondson et al. 2000).

If decomposers are acceptable, but less desirable prey, they can sustain predator populations when pest pressures are low, but not interfere with control when pest populations increase. Unfortunately, it is difficult to predict how valuable decomposers are for pest management because interactions between generalist predators and decomposers are poorly documented. The generalist predator community in Pennsylvania field crops is dominated by carabid beetles and spiders. Some carabid beetles reject earthworms and isopods (Best and Beegle 1977), while other species readily accept earthworms or juvenile millipedes as alternative prey (Brunke et al. 2009, Symondson et al. 2000). Certain spiders have been shown to heavily rely on decomposers, namely collembolans (Agusti et al. 2003). For millipedes especially, predation has rarely been studied, and most references to predators are based on casual observation and anecdotes (Sierwald and Bond 2007). More studies on millipede-predator interactions, and the defensive mechanisms involved, can aid our understanding of the role of decomposers in pest management.