Recently, we published a paper with the title “Comparative analyses of glycerotoxin expression unveil a novel organization of the bloodworm venom system” and with this “behind-the-paper-style” blog I want to give some insights how we came to our conclusions. The paper was published by Richter et al. (2017) in BMC Evolutionary Biology and is open access.
Venoms and venom systems evolved many times independently across the animal tree of life. That spiders or snakes can be venomous is well-known. Also, the venom of the iconic cone-snails has been investigated in detail. However, venomous annelids have been (mostly) neglected so far. There are around 20,000 annelid species, but only few of them have been convincingly shown to be venomous. E.g., fireworms (Amphinomidae) can burn like hell if you touch them, which is due to an inflammation-inducing substance called complanine. However, this is passively delivered when their chaetae break and therefore these animals should be regarded as poisonous (which still make them an interesting target for studying the content and evolution of their toxins). In contrast, venomous are those animals which actively deliver their toxin cocktail,. e.g., for predation, defense or competition. Only glycerids and leeches are convincingly demonstrated to represent venomous annelids (von Reumont et al. 2014). For several other annelids (e.g., chrysopetalids, polynoids or sigalionids) it has been speculated that they might be venomous, however, our own dissections could not reveal convincing evidence of venom glands or an apparatus to deliver the venom in several investigated taxa. Nevertheless, different groups of scale-worms (e.g., Pisione spp., Pholoe spp.) remain good candidates to include venomous species, but desperately need to be studied in detail. And I also would not be surprised if more examples from other annelid taxa will be discovered in the future.
With currently ~50 recognized species Glyceridae or bloodworms (Fig. 1) are a rather small annelid taxon. However, its likely that the true diversity has been underestimated due to their anatomical uniformity.
Fig. 1: Some glycerids, which not surprisingly are a popular fishing bait.
Glycerids are distributed worldwide and easily recognizable by an eversible pharynx that possesses four cross-arranged jaws (Fig. 2), which are associated with a venom gland. Or at least, that was what everybody thought these structures would be…
Fig. 2: All glycerids possess four cross-arranged jaws, which are associated with gland-like structures. Modified after Ehlers (1868).
Animal venoms received much attention due to their potential for the development of pharmaceutical drugs or medical tools. Toxins are often highly specific in their mode of action. Many of them activate or block specific channels at the neuromuscular junction. E.g., some conotoxins of the venomous cone snails (Fig. 3) specifically block N-type voltage-gated calcium channels. These toxins became the posterchild for venom research, as they were successfully used to develop a drug called zicotonoide or prialt, which is a pain reliver that is more powerful, but less addictive then opioids. Several other conotoxins with different cellular targets are currently investigated for their use in drug development (Bingham et al. 2010). Also the venom of glycerids contained some pharmaceutically interesting toxins. Best known is a neurotoxin called alpha-glycerotoxin, which has been demonstrated to specfically activate the already mentioned N-type calcium channels (Meunier et al. 2002; Schenning et al. 2010). As this mode of action is also dose-dependend and fully reversible, a potential use of glycerotoxin as a drug or medical tool in neuroscience became obvious. All work on the function of this neurotoxin has been done on the protein level, but no nucleotide sequence was available so far. However, working on a DNA-level is neccessary for recombinant expression of this protein by bacteria (or other systems for protein production), to make it available in large amounts.
Fig. 3: Conus bullatus shell (left) and living animal catching prey (right). Reprinted from Hu et al. (2010).
My personal interest in the venom evolution of glycerids started back in 2005. I was inspired by the work of Thomas F. Duda and Stephen R. Palumbi, who demonstrated remarkable convergence in the evolution of conotoxin sequences related to the feeding ecology of cone snails (e.g, Duda & Palumbi 2004). After reading their paper I directly thought about planning a similar study for glycerid annelids. There were only two problems. There was no glycerotoxin sequence available to work with – and we also do not know much about the feeding ecology of glycerids (Jumars et al. 2015). To start with the sequence data, I contacted Giampetro Schiavo (now UCL, formerly Cancer Research UK), who studied the impact of glycerotoxin on neurotransmitter release. Luckily, in the lab of Giampietro they already sequenced short snippets of the glycerotoxin protein sequence, which could be used to find a match in DNA data. They tried themselves with degenerated primer, but did not succeed. As I was involved in a project generating transcriptomic data for a phylogenomic study of annelid relationships, I thought having a look at some glycerid transcriptome data might give a better chance finding the glycerotoxin sequence. However, this was back in 2007/2008 when Sanger sequencing was still dominating and next generation sequencing (NGS) was just starting to become powerful and also available to a broad community. At least I got some 454 data (a NGS technique which is already retired…), but I could not find any good candidates for the glycerotoxin gene in my data. I started a new try when I moved to Leipzig University in late 2009. This time, with help from Matthias Meyer‘s group at the Max Planck Institute for Evolutionary Anthropology we were able to use the much more powerful Illumina sequencing technique. Instead of analysing a few thousand sequences (if at all), we were now able to analyse several million sequences. My master student Joerg Hetmank finally succeeded in screening different RNAseq libraries and we found the glycerotoxin sequence. Interestingly, we found it in a library from an anterior end – not in the so-called venom glands…
First, we only recovered a 1000 bp contig which raised more new questions then solving old ones. Most known animal toxins have been shown to evolve from gene families which ancestrally have a different function. Such mode of evolution by gene duplication followed by neo-functionalization or sub-functionalization has been demonstrated for example in snakes or spiders (e.g., Fry et al. 2009). However, comparisons of our sequences with those in databases did not reveal any annotated domains. From the protein work we also knew that the gene must be much bigger. Sandy Richter started to work on this topic in her master thesis and subsequently in her PhD-thesis. Using more and deeper transcriptome sequencing data she was able to recover the complete neurotoxin. Or more precise, neurotoxins, as glycerotoxin obviously is a gene family and several copies can be found in the genome of Glycera tridactyla. Full length transcripts have lengths of ca. 3600 bp and this time at least some short domains could be identified at the N-terminal end (Fig. 4), which seem to be important to bind at the calcium channels.
Fig. 4: Annotation of the glycerotoxin gene, showing some known domains at the N-terminal end of the protein. However, the largest part of the protein sequence still remains un-annotated.
Sandy also started to do some morphological analyses of the venom system of glycerids and wanted to see if this neurotoxin shows some interesting expression patterns in the venom gland. Using in situ hybridization techniques, she started to design probes which will bind to the RNA of this gene and theirby help to reveal the place of expression. As the whole animal is to big for this methodology, she dissected the venom glands to work with. To our surprise, the results of this experiments showed that the neurotoxin was obviously not expressed in the so called venom glands – but within lobes attached to them (Fig. 5).
Fig. 5: (a) Overview of an inverted glycerid pharynx with four putative venom glands each connected to a jaw,. Note the four corresponding lobes (three of them marked by an arrow) connect to these “glands”. (b) Fluorescence in situ hybridization (FISH) revealed a clear glycerotoxin signal only in the lobes. (c) Consistent results are found when using anti-glycerotoxin antibody. Venom stored within the “glands” is likely washed out.
One initial interpretation of these results is that we think that the lobes are the place of glycerotoxin production, whereas the “glands” are rather reservoirs for the venom. This would also explain our initial failure of finding glycertoxin in the “gland” transcriptome. However, we wanted to test this idea with another approach, also involving more individuals. Therefore we designed a real-time PCR experiment, where we quantified the relative expression level of glycerotoxin in three different tissue types: lobes, “glands, and body wall (Fig. 6)
Fig. 6: Tissues analyzed with real-time PCR: (a) lobes, “gland” and (b) body wall.
The real-time PCR experiments were carried out by Stephan H. Drukewitz for his master thesis. For this experiment, RNA had to be extracted for all tissue types for all individuals and subsequently transcribed to cDNA for the real-time PCR: As we already had an idea about the intron-exon structure of the glycerotoxin gene based on Sandy’s work, we were able to design primers where at least one oligonucleotide was covering a region which is interrupted by an intron in genomic DNA: This helped to avoid false signal due to contamination by genomic DNA. The results were clear-cut and did not need much statistical interpretation. A high expression of glycerotoxin is found in the lobe tissue of animals, low expression in the “glands” and basically no expression in the body-wall. We found that different paralogs of the glycerotoxin genes are on average 800x higher expressed in the lobes than in the “glands” (Fig. 7).
Fig. 7: Relative expression of glycerotoxin in the lobe tissue (lobes) vs. “gland” tissue.
In summary, our results clearly demonstrate that the pharyngeal lobes are the main location of neurotoxin expression. Now we basically have two possibilities how to interpret the glycerid venom system. Either the lobes are the sole place of venom production and the “glands” are only reservoirs, storing the venom. Or, glycerids have a venom system with two subunits, which may produce different toxins. Something similar has been reported for cephalopods, which bear an anterior and a posterior gland (Fry et al. 2009). To answer this question, the protein content of milked glycerid venom (thanks god, my collaborator Ron Jenner from the NHM in London, who was also involved in this study, is a skilled glycerid venom milker) should be analysed in detail in the future, which then can be matched with transcriptomic data from the two different tissue types. But of course several other questions remain open: What is the evolutionary origin of the glycerotoxin gene family and how it is distributed in glycerids? How exactly acts the glycertoxin and can we produce it synthetically? Do all glycerids have a similar acting venom or are there huge diffenrences correlating with their feeding biology? This last questions brings us back to my initial inspiration to investigate this topic and reminds me that we do not know much about the ecology of glycerids. Are all species carnivorous predators – or are some detrivores? Hopefully the future will bring answers to some of these questions.
Sometimes science takes time, a luxury which is often not available in our academic environment. From the first idea doing this study (2005) till its publication (2017) it took altogether 12 years. Three master theses and one Phd-thesis contributed to this study in my lab alone and I am also happy to acknowledge support from the German science foundation (DFG). Moreover, without the pioneering work by Giampietro Schiavo and Frederic Meunier, who made protein sequence snippets and an antibody available, this study would have been impossible from the start. Besides the people I mentioned so far, I also want to note that Conrad Helm, Lars Hering, Lahcen Campbell and Eivind Undheim contributed to this study. Finally, its done!