Transgenesis enables mapping of segmental ganglia in the leech Helobdella austinensis

The stereotypy of individually identified neurons has made invertebrate species with simple nervous systems useful for understanding the operational principles of neurons and neural circuits. In such nervous systems, landmark neurons, individually identified by location, morphology (especially size) and electrophysiological properties, can be recorded and manipulated to characterize their roles in neural circuits. Further, when the stereotypy extends beyond species boundaries, homologous neurons from different species can be compared to study the evolution of neural circuits. Assessing the stereotypy of neuronal elements beyond the easily recognizable landmark neurons has been difficult because of technical limitations, however. Closing this knowledge gap will help us better understand the plasticity and robustness of these simple nervous systems.

Among the classic invertebrate models, leeches are clitellate annelids, i.e. segmented worms in the superphylum Spiralia. The leech body plan arises from 32 segmental primordia (plus additional non-segmental tissues), organized into four fused rostral/head segments (R01–R04), 21 midbody segments (M01–M21) and seven caudal/tail segments (C01–C07). The leech nervous system is also segmentally organized; each midbody segment bears a discrete ganglion that is homologous to those in other segments, and the fused head and tail segments contain ganglionic masses composed of fused segmental ganglia (Shankland and Savage, 1997).

Two leech families, Hirudinidae and Glossiphoniidae, have been used as models for neurobiology and embryology, respectively, since the 19th century (Retzius, 1891, 1898; Whitman, 1878, 1887). Hirudinid species (e.g. Hirudo medicinalis and Hirudo verbana) are well suited for studying cell physiology and the neural substrates of behavior; their 32 segmental ganglia comprise identified neurons that are accessible for intracellular recording (Kristan et al., 2005; Szczupak, 2023; Wagenaar, 2015). Leech nervous systems are also of interest for comparative analysis for two reasons. First, the leech nervous system (∼400 neurons per ganglia; ∼12,000 in the CNS) is intermediate in complexity between those of more intensively studied ecdysozoan models, i.e. Caenorhabditis elegans (302 neurons total in the hermaphrodite) and Drosophila melanogaster (∼100,000 neurons in the brain alone). Second, as spiralians, leeches represent a group that has been evolving independently of the ecdysozoans for ∼600 MY; thus, comparisons between leeches and other, distantly related organisms (e.g. Drosophila and vertebrates) should help to illuminate aspects of neurobiology and behavior that are deeply conserved and also those that are evolutionarily plastic.

Unfortunately, the early embryos of Hirudo are small and have never been cultured out of the cocoon, which makes them intractable for the intracellular injection of reagents for cell lineage tracing or genetic manipulation. For such work, glossiphoniid species, e.g. Helobdella austinensis (Fig. 1Ai,ii) are useful because their embryos are easily cultured in the laboratory and because the large, identifiable blastomeres in their early embryonic stages (Fig. 1B–D,F) are amenable to a variety of experimental approaches, including intracellular injection (Kuo et al., 2020; Weisblat, 2022; Weisblat and Kuo, 2009, 2014). However, the small size of its adult nervous system limits physiological approaches to studying it.

To address those limitations and to establish groundwork for future developmental and neurophysiological studies, we have developed efficient and lineage-specific transgenesis in Helobdella embryos. For this purpose, we microinjected individual blastomeres of the early embryo with a cocktail of a Minos transposon-bearing plasmid and Minos transposase mRNA. Injecting germline precursors allowed us to generate lines of Helobdella that stably express selected transgenes. Combined with the judicious identification of a pan-neuronal enhancer, we have used Minos transgenesis to determine the neuronal content and variability in ganglia of the ventral nerve cord in Helobdella.

Helobdella austinensis (Kutschera et al., 2013) embryos were obtained from a long-term laboratory colony, staged and cultured all as described elsewhere (Weisblat and Kuo, 2009). Plasmids were pressure-injected into identified blastomeres at stages 4c–7 (DM blastomeres, OP proteloblasts or M teloblasts), using the same procedures as for microinjecting HRP and derivatized dextran lineage tracers (Weisblat and Kuo, 2009). To optimize survival, injected embryos were frequently cultured through stage 8 in HL saline containing roughly 1% commercial penicillin−streptomycin solution (Gibco).

In the present experiments, injected embryos (F0 animals) were screened under a fluorescence dissecting microscope to visualize neuronal expression of fluorescent protein (FP) driven by the Hau-elav upstream regulatory elements at stages 9–10, when embryonic mobility was still minimal and yet neuronal differentiation was well under way. FP-expressing individuals from injected clutches were cultured further to the end of stage 11 (yolk exhaustion) and then hand-reared to reproductive maturity on a diet of small physid or planorbid snails. These F0 animals were allowed to interbreed; the resulting clutches of F1 animals comprised mainly clutches with no FP expression and a few clutches of mixed expression and non-expression. FP-expressing F1 animals were selected to interbreed. As the occurrence of FP-expressing F1 animals in each clutch is typically low (0–29%) and because Helobdella juveniles survive better when cultured as groups, we often pooled FP-expressing F1s to improve survival rates and preserve transgenes within the population. The resultant F2 clutches usually contained a higher proportion of FP-expressing animals; this allowed us to select and grow cohorts of FP-expressing F2s from individual clutches in isolation to establish inbreeding lines. We scored the proportion of FP-expressing animals among the F3 clutches to estimate the inheritance pattern of transgenes. F3 clutches with 100% transgene expression generally gave rise to progeny consistently expressing the FP reporter.

Four elav homologs were identified from the InterPro Helobdella robusta proteome dataset (accession number: UP000015101) using the query IPR006548, an InterPro ID that structurally defines the Elav/Hu family members. The amino acid sequences of the four H. robusta elav proteins retrieved from the InterPro database were used to perform TBLASTN searches of gene models in the H. robusta genome database (https://mycocosm.jgi.doe.gov/Helro1/Helro1.home.html) and H. austinensis transcriptomes to obtain their nucleic acid sequences. For each of these four Hau-elav genes, an ortholog was identified in H. robusta.

Phylogenetic analysis of Elav proteins from selected invertebrate species was performed using additional sequences retrieved from the EnsemblMetazoa database. The amino acid sequences were aligned using the Muscle algorithm (Edgar, 2022). Maximum-Likelihood tree search was performed using IQ-TREE (Nguyen et al., 2015); the best-fit substitution model was identified by ModelFinder (Kalyaanamoorthy et al., 2017) and automatically implemented during the tree search sessions using the optional argument -m MFP in IQ-TREE (http://www.iqtree.org/doc/); 1000× approximated bootstrapping was performed using the ultrafast bootstrap algorithm (Hoang et al., 2018). The plasmid templates for synthesizing antisense riboprobes of the H. austinensis elav genes (Hau-elav1, Hau-elav2, Hau-elav3 and Hau-elav4) were produced by PCR cloning of cDNA fragments. The primer pairs for amplifying these cDNA fragments are shown in Table S1. Probe synthesis and whole-mount in situ hybridization were performed as previously described (Weisblat and Kuo, 2009).

Genomic organization for both Hau-elav4 and Hau-cif7 was reconstructed by mapping the H. austinensis genome assembly with the EST-supported H. robusta gene models and H. austinensis transcripts recovered from the transcriptomes. The genomic organization of ortholog pairs is conserved between H. robusta and H. austinensis. The upstream genomic fragments of Hau-elav4 and Hau-cif7 were PCR amplified using primers listed in Table S1 and cloned into the pJET vector. The resulting plasmids contain a 3552 bp fragment corresponding to −3549 to +3 bp relative to the Hau-elav4 translation initiation site and a 4279 bp fragment corresponding to −4255 to +24 bp relative to the Hau-cif7 translation initiation site, respectively. These two plasmids served as the starting materials for subsequent reporter constructs.

The base of the reporter constructs was modified from the pMi{3XP3-dsRed} plasmid (Pavlopoulos and Averof, 2005). Briefly, the 3XP3-dsRed insert was excised by restriction digestion with BamHI, followed by self-ligation. The resulting pMinos plasmid contains several unique restriction sites that can be used to insert DNA fragments between the left and right Minos elements.

A DNA fragment containing the SV40 polyadenylation signal was excised from the pCS107 vector by double digestion with XhoI and AscI and inserted into pBSMNEF1P (Gline et al., 2009), resulting in pEF-SV40. Next, the DNA fragment containing the 2270 bp Helobdella triserialis EF1α upstream sequence and the SV40 element was excised from pEF-SV40 by double digestion with BamHI and SacII and then inserted into the pMinos plasmid to yield pMi{Htr-ef1a-sv40}. ClaI and XhoI sites were introduced into the 5′- and 3′-ends of nGFP, H2B:GFP and H2B:mCherry coding sequences by PCR amplification from pCS2+nGFP (Zhang and Weisblat, 2005), pCS107 H2B:GFP (Gline et al., 2009) and pCS107 H2B:mCherry, respectively. The restriction enzyme-digested amplicons were then inserted between the ClaI and XhoI sites of pMi{Htr-ef1a-sv40} to generate pMi{Htr-ef1a-ngfp} and pMi{Htr-ef1a-h2b:gfp}. pC107 H2B:mCherry was created by replacing eGFP coding in pCS107 H2B:eGFP with the mCherry coding sequence. The primer pairs for amplifying DNA fragments encoding nGFP, H2B:GFP and H2B:mCherry are listed in Table S2.

The tissue-specific reporter constructs (pMi{Hau-elav4-ngfp}, pMi{Hau-elav4-h2b:mcherry} and pMi{Hau-cif7-ngfp}) were produced using the NEBuilder HiFi DNA Assembly kit (New England Biolabs, NEB). The pMi base plasmid was linearized by double digestion with SpeI and SacII. The upstream DNA fragments were PCR amplified from their respective plasmid templates. The DNA fragments encoding the nGFP-SV40 and H2B:mCherry-SV40 reporters were amplified from the pCS2+ nGFP and pCS107 H2B:mCherry. The combinations of primer pairs and plasmid templates for generating DNA fragments used in the DNA assembly reactions are listed in Table S3. The resulting pMi{Hau-elav4-ngfp} and pMi{Hau-elav4-h2b:mcherry} plasmids contain a 2587 bp DNA fragment upstream of the start codon of Hau-elav4, followed by an nGFP-SV40 or H2B:mCherry-SV40 reporter, respectively. The pMi{Hau-cif7-ngfp} plasmid contains a 4254 bp fragment upstream of Hau-cif7, followed by an nGFP-SV40 reporter.

Genomic DNA of transgenic or wild-type leeches was individually extracted using a ZR Genomic DNA Tissue Microprep Kit (ZymoResearch). Following quantification, genomic DNA was restriction digested with HindIII (NEB) and circularized by self-ligation. The cloned Hau-elav4 upstream region contains a single HindIII site near the 5′ end, and the frequency of HindIII sites in the Helobdella genome was estimated to be one per 2916.4 bp on average. The concentrations of genomic DNA (0.1 ng μl⁻¹) and T4 DNA ligase (0.04 U μl⁻¹) and the reaction temperature (16°C) were kept low to suppress the formation of heteroduplex DNA. After ligation, 1 ng of self-circularized genomic DNA was used as a PCR template and amplified with a pair of inversely oriented primers. The forward primer annealed to the SV40 polyadenylation signal region, and the reverse primer annealed to a region immediately downstream of the HindIII site in the cloned 2587 bp Hau-elav4 upstream DNA fragment. The sequences of this primer pair are given in Table S4. This amplicon is designed to capture endogenous genomic DNA fragments adjacent to the Minos right arm at the insertion site. Amplicons were analyzed by agarose gel electrophoresis to determine the number of insertions in each individual. Selected amplicons were sequenced to confirm genomic DNA integration and to determine the genomic location of the insertion sites.

For the initial evaluation of the Hau-cif7 and Hau-elav4 reporters, the embryos injected with the transgenic cocktail were raised to stage 10 and stage 11, respectively, and fixed at 4°C overnight in a fixative containing 4% formaldehyde and 0.5× PBS. The fixed specimens were counterstained with Hoechst DNA stain. For stage 10 ganglia, the germinal plate was dissected off the yolk, flattened, and mounted in 80% glycerol. Images of stage 11 ganglia were acquired from whole-mount specimens cleared with the DeepClear protocol (Pende et al., 2020).

For high-resolution analysis of segmental ganglia, three animals were dissected along the dorsal and ventral midlines, fixed in 4% formaldehyde overnight, rinsed in PBS, and cleared in 40%, 60% and 80% glycerol. Finally, they were mounted in 80% glycerol for imaging. Our aim was to count neuronal nuclei in the 17 midbody segmental ganglia M02 to M18 for each of the three individuals. Limitations imposed by the dissection protocol made it difficult to access segments anterior to M02 and posterior to M18. Of these 51 target ganglia, three ganglia could not be imaged (either at all or in part) because of obstructions arising during dissection and mounting. Thus, our final dataset consisted of 48 complete ganglia (288 packets). Ganglia were imaged using a 20× objective on a Zeiss LSM 900 confocal microscope, ventral side up; mCherry fluorescence was excited by a 561 nm laser. Whole-ganglia images were obtained using a Z-spacing of 2 µm. Individual packets were captured as 8-bit grayscale images (resolution 2048×2048 pixels), using a digital zoom between 2.5× and 3.8× with a Z-spacing of 0.63 µm. Data were acquired as CZI files and transformed to TIF stacks for analysis.

To optimize resolution, we imaged ganglia on a packet-by-packet basis. Confocal images were processed using ImagePro Plus (v6.3, Media Cybernetics Inc., Rockville, MD, USA). Stack images were used for generating 2D or 3D images. 2D images were used to measure the area (area of the Z-projected nuclei), mean diameter, roundness [perimeter²/(4π×area); roundness=1 is circular], perimeter, mean intensity (average intensity of all the pixels in the object) of the fluorescent nuclei and the integrated optical density/intensity (IOD) which reports the average intensity/density of each object (nuclei) multiplied by the area of that object (Portiansky, 2018). For this purpose, the Extended Depth of Field [EDF, also known as Extended Focus Image (EFI; Olympus cellSens) or Z-projection in MIT Image (FIJI)] option was applied using the Max intensity focus analysis option. Nuclei in calibrated images were manually circumscribed and simultaneously counted and measured. To test the accuracy of the counting procedure, the neurons identified in selected packets were checked independently and found to be identical. Raw data are provided in Dataset 1.

3D images were used as a model to identify nuclei in 2D projected images and to measure volume parameters of some packets. The 180 deg rotated rendered 3D images were used to detect nuclei hidden behind large neurons on the dorsal side of the ganglion (Movie 1). Volume parameters (volume, surface area, diameter, sphericity) were calculated using the 3D Constructor plug-in algorithm of the image analysis software. Data were exported to a spreadsheet for statistical analysis.

A previous application of transgene-based lineage tracers in Helobdella used a plasmid in which a pair of I-SceI recognition sequences flanked a transgene cassette consisting of a 2270 bp genomic DNA fragment upstream of the coding sequence of Helobdella triserialis elongation factor-1-alpha (Htr-ef1a) and a FP reporter (Gline et al., 2009). Injecting this plasmid alone resulted in transient ubiquitous reporter expression with low mosaicism among the progeny of the injected cell in F0 embryos, making it useful for lineage tracing in the leech embryo, but co-injecting this plasmid with the transposase (I-SceI meganuclease) yielded no stable integration.

As an alternative to the meganuclease approach, constructs based on the Minos transposon have been used to insert transgenes into the host genome in various arthropod and chordate species (Pavlopoulos et al., 2007). To see whether the Minos system is useful in H. austinensis – a spiralian species that is phylogenetically distant to arthropods and chordates, we generated Minos-based transgene constructs (pMi{ef1a-ngfp} and pMi{ef1a-h2b:mcherry}) by transferring the transgene cassette from the previously made I-SceI plasmids into a plasmid containing the Minos transposable elements (Fig. 1E). We then injected early blastomeres with a transgenesis cocktail of the plasmid and mRNA encoding Minos transposase mixed with a fluoresceinated dextran tracer (Fig. 1Fii) and monitored the transgene reporter expression in the injected F0 individuals and their descendants. As expected, we observed uniform expression of the transgene at early times following injection, reflecting plasmid-driven expression (Fig. 1G–I; Gline et al., 2009). Expression at later stages became mosaic, as unintegrated plasmids were degraded, while expression from the integrated transgene persisted (Fig. 1J).

To achieve germline transmission of the transgene, we injected the transgenesis cocktail into embryonic germline precursor cells, i.e. M teloblasts or their precursors in the D quadrant lineage (Cho et al., 2014; Kang et al., 2002). In one experiment, the D′ blastomere was injected in 8-cell embryos (stage 4a). Of 42 injected embryos, 27 (64%) developed normally to adulthood, 22 of which showed readily visible fluorescence. Most of the expected progeny cells of the injected blastomere expressed FP; mosaicism was evidenced mainly by differences in the level of expression, especially within the epidermis, segmental ganglia and testes sacs (Figs 2A,B). In some cases, mosaicism allowed us to infer the timing of integration events relative to the stereotyped cell division patterns of the injected lineage. For example, Fig. 2C shows a young F0 adult grown from an embryo in which the transgenesis cocktail was injected into mesodermal precursor DM shortly before its equal division to form the bilateral pair of M teloblasts; evidently, the transgene was incorporated into one M teloblast but not the other, resulting in unilateral mesodermal expression of the transgene. To establish a stable transgenic line, the transgene-expressing F0 individuals selected from the experiment described above were grown to adulthood and allowed to interbreed; 20–40% of the animals in the F1 generation exhibited seemingly ubiquitous transgene expression (Fig. 2Di,ii). However, it is noteworthy that reporter expression was first observed only in micromeres during stage 5 (not shown) and only became ubiquitous in later developmental stages in the stable lines, suggesting that the Htr-ef1a element is not sufficient to drive maternal and early embryonic expression of the reporter.

Initiating transgenic lines by breeding cohorts of independently injected F0 individuals means that the resulting lines should be heterogeneous with respect to the number of integrated transgenes and their chromosomal locations. Landing the transgene in different locations is expected to result in subtle differences in expression from, for example, differences in the local enhancer landscape. This assumption is supported by observed differences among individuals of a breeding line bearing the ef1a-h2b:mcherry transgene, in terms of the overall brightness, and in the relative brightness of various cell types even among size-matched individuals (Fig. S1).

Having demonstrated the feasibility of generating transgenic Helobdella lines using the ef1a enhancer, we sought to identify enhancer elements for genes expressed pan-neuronally, to implement transgenesis approaches for studying the nervous system of Helobdella. A previous analysis of nine cytoplasmic intermediate filament (cif) genes in Helobella revealed that Hau-cif7 is strongly and specifically expressed in the developing nervous system (Kuo and Weisblat, 2011). In the present work, a 4279 bp fragment upstream to the Hau-cif7 translation start site was amplified by PCR and then cloned into the Minos transgenesis vector, yielding pMi{Hau-cif7-ngfp} (Fig. 3A). To determine the cell-type specificity of reporter expression, we injected an OP proteloblast with the transgenesis cocktail and found that the Hau-cif7 enhancer drove transgene expression in only 3–5 cells in each hemiganglion (Fig. 3B). The number of O- and P-derived neurons in a hemiganglion is much greater than the observed 5 (Kramer and Weisblat, 1985; Weisblat et al., 1984), implying that the cloned Hau-cif7 upstream region does not promote pan-neuronal expression. Furthermore, in stage 11 embryos, the GFP-positive cells were localized on the surface of ganglionic glial packets, whereas the interior of glial packets, where the ganglionic neurons are expected, was devoid of GFP-positive cells (Fig. 3Ci,ii). Therefore, based on their number and distribution, these GFP-positive cells are likely ganglionic packet glial cells rather than neurons. Thus, this Hau-cif7 regulatory element does not satisfy our requirements for this work.

The mRNA-binding protein Elav is required for neuronal differentiation in Drosophila and elav homologs have proven to be broadly conserved markers for post-mitotic neurons in various metazoan species (Colombrita et al., 2013; Pascale et al., 2008). The Helobdella genome contains four elav paralogs (Fig. 3D). Expression analysis by in situ hybridization indicates that, despite minor differences in expression patterns (Fig. 3E–H), three of the four Hau-elav genes are expressed in segmental and supraesophageal ganglia of developing embryos. Orthologs of these three elav genes are also all expressed by neurons in the adult ganglia of H. verbana (Heath-Heckman et al., 2021). Hau-elav4 was selected for enhancer cloning on the basis of its compact genome organization. A 2588 bp fragment upstream to Hau-elav4 (Fig. 3I) was amplified by PCR and cloned into the pMinos transgenesis vectors, yielding pMi{Hau-elav4-ngfp}. Injection of pMi{Hau-elav4-ngfp} transgenesis cocktail into OP proteloblast gave robust expression in previously identified neuronal elements and not in other cell types; furthermore, faintly labeled neurites were visible in advanced developmental stages (Fig. 3J,K). Thus, the cloned upstream element of Hau-elav4 contains an enhancer capable of driving gene expression in neurons specifically.

To determine whether this upstream DNA fragment of Hau-elav4 drives pan-neuronal reporter expression, we examined the transgenic F1 progeny of F0 parents injected as embryos with a transgenesis cocktail containing pMi{Hau-elav4-ngfp} into a germline precursor cell, the M teloblast. In the successfully transgenic F1 individuals, we detected transgene expression in previously identified neuronal elements including segmental ganglia of the ventral nerve cord, the supraesophageal ganglion, and presumptive peripheral sensory neurons of the prostomial lip and the body wall sensilla (Fig. 4A,B). Therefore, the Hau-elav4-driven reporter expression is likely pan-neuronal.

To establish transgenic lines, we crossed transgene-expressing F1 individuals among themselves, as described above. It has been previously reported that the Minos system regularly results in multiple insertions in the host genome (Pavlopoulos and Averof, 2005). Assuming that many chromosomes would have undergone one or more transgene insertions, we anticipated that F2 cohorts derived from transgenic F1 parents would be nearly 100% positive for transgene expression. Contrary to this expectation, however, none of the F2 cohorts had such high frequencies of transgene expression, indicating that the majority of F1 individuals may have only one or two insertions per genome and were heterozygous at the insertion loci. If so, F2 cohorts derived from sibling heterozygous F1 parents should contain some homozygotes. To test this prediction, we crossed sibling F2 leeches that were positive for transgene expression and scored the rates of transgene expression in the resulting F3 cohorts: 14.68% (48/327) of F3 cohorts were found to be 100% positive, confirming that some of the F2s were indeed homozygous. The rates of transgene expression among F3 cohorts generally fell between 0.6 and 1.0, which is also consistent with the predicted inheritance pattern of 1–2 transgene insertion loci (Fig. 4C). Overall, the expression of transgenes in our transgenic lines approximates the Mendelian ratio of a single-locus dominant allele. To verify the number of inserted transgene loci directly, we performed inverse PCR genotyping on transgenic animals taken from selected true-breeding lines. One out of the seven lines examined contained two loci, and the remaining six lines contained only a single locus (Fig. 4D). Therefore, despite having observed high success rates of transgene expression following injection of a transgenesis cocktail, which suggests a high efficiency of insertion, the Minos transposon induced an unexpectedly small number of transgene insertions in the Helobdella genome. The factor(s) limiting the number of transgene insertions is currently unknown, but the ease of obtaining homozygotic single-locus transgenic animals is an unexpected advantage of using Minos-mediated transgenesis to study Helobdella biology.

As described in the Introduction, each midbody segment contains a distinct segmental ganglion, numbered M01–M21 as are the segments themselves (Fig. 1Aii). Within each ganglion, neurons are monopolar, with cell bodies distributed among six anatomically distinct packets: two packets make up an anterior and posterior pair straddling the ventral midline and are designated AVM and PVM, respectively; four more packets constitute left/right pairs of anterior and posterior lateral packets and are designated LAL, RAL, LPL and RPL, respectively (Fig. 5A). Within each packet, nerve cell bodies and their proximal neurites are enveloped by processes of a giant packet glia cell (Coggeshall and Fawcett, 1964). Distally, the neurites extend into the central neuropil, where they are enveloped by two other giant glia (Coggeshall and Fawcett, 1964). In addition to the giant packet and neuropil glia, ganglia also contain numerous other non-neuronal cells, including muscle cells, sheath cells and microglia (Coggeshall and Fawcett, 1964; McGlade-McCulloh et al., 1989). Prominent longitudinal axon tracts traverse the dorsal aspect of each ganglion, continuous with the interganglionic connective nerves (Fig. 5A). Sensory and motor neurons reach the periphery through left and right segmental nerves.

In leech ganglia, neurons are identified in part by their size and location within the ganglia, and within specific packets. For example, a pair of large serotonergic neurons, known as Retzius (Rz) cells, are readily detected by their serotonin content throughout clitellates, i.e. oligochaetes and leeches (Koza et al., 2006). To assess the neuronal inventory of segmental ganglia in H. austinensis, we determined the absolute numbers and sizes of nuclei in mid-body ganglia from three randomly selected individuals from a Hau-elav4-h2b:mcherry transgenic line. We used this marker for the analysis on the assumption that the elav enhancer is expressed in all ganglionic neurons, and not in glia or sheath cells, as described above (Figs 3 and 5Bi,ii). Moreover, an intensity histogram of all 15,317 observations revealed very few at either extreme (Fig. S2A), suggesting that we are capturing all the labeled nuclei. And finally, the histone2B:mCherry fusion protein is tightly localized to chromatin of neuronal nuclei (Gline et al., 2009), making it easier to distinguish closely apposed cells (Fig. 5Bii). For each of 48 ganglia (288 packets) drawn from segmental ganglia M02 to M18 in three individual leeches (L1, L2, L3), we counted mCherry-positive neuronal nuclei on a packet by packet basis (see Materials and Methods for details).

Based on the evidence of neural determinacy for the leech nervous systems (Muller et al., 1981), we expected that Helobdella ganglia and packets would comprise nearly invariant numbers of neurons with stereotyped size and spatial distribution, subject to possible segment-specific differences. In fact, however, the ganglia we examined exhibited substantial variability. Homologous ganglia typically exhibited a considerable range in neuronal counts among the three specimens, and one of the specimens had consistently lower neuron counts overall (Fig. 5C,D; Dataset 1), giving rise to a statistically significant difference among the three specimens (Kruskal–Wallis test, P=3.7e−05). Pairwise comparisons among the three specimens (Dunn's test) showed that L1 differed from L2 (P=2.4e−05) and from L3 (P=0.01), while L2 and L3 were not significantly different from each other (P=0.3).

The total number of nuclei counted in the different ganglia varied both within and among the specimens examined whether comparing positionally homologous segments in the three specimens, serially homologous segments in each individual, or even ganglia of reproductive (M5 and M6) and other segments (Fig. 5C,D). In the face of this unexpectedly high variability, no significant differences were observed among the neuronal complements of the 17 midbody ganglia examined here (Kruskal–Wallis test, P=0.4; Fig. 5C). Overall, the 48 ganglia spanned a range from 249 to 382 nuclei (319±26, mean±s.d.) with a coefficient of variation of 0.08 (Fig. 5C,D).

The distribution of neurons among the six packets was also variable (Fig. 5E). However, despite the fact that the neuron counts for each of the six packets varied by as much as 2-fold, we observed that the midline packets (AVM and PVM) contained significantly fewer nuclei than the lateral packets (LAL, LPL, RAL and RPL; Fig. 5E). The mean coefficient of variation for the number of neurons in the different packets was around 0.15 (0.15±0.01). Thus, the distribution of neurons per packet varied more than the distribution of neurons per ganglion. This supports the possibility that neuronal cell bodies arise in somewhat predictable, but not rigidly fixed, positions during development, and that neuronal cell bodies lying between two packet glia end up being distributed stochastically among their respective packets during development. Consistent with this, we observed that while AVM contains two large, presumptive Rz neurons in most of the ganglia in L3 (Fig. 6), the two large cells are distributed between the AVM and PVM packets in two ganglia (M04 and M06), and in ganglion M11, the AVM packet contains three large cells. Rz abnormalities were also observed in four ganglia of L2 (M02, M06, M08 and M17; data not shown), but all 15 ganglia that we were able to image from L1 were normal. Thus, a total of 7 of the 48 ganglia examined exhibited atypical distribution of presumptive Rz cells.

As alluded to in the preceding paragraph, another feature used to identify individual neurons in the leech ganglia is a pronounced heterogeneity in cell body size, presumably reflecting differences in the cytoplasmic volume required to meet the needs of cells with differing biosynthetic/metabolic activities, and possibly driven by differences in ploidy among post-mitotic neurons (Lasek and Dower, 1971; Yamagishi et al., 2011). For simplicity, we used the area of the 2D projection of the labeled nuclei as a proxy for volume, after first testing the validity of the correlation between projected area and measured volume in a single ganglion (Fig. S2B–D; see Materials and Methods for details).

The projected area of the nuclei we counted varied over a >150-fold range (Fig. S3, Dataset 1). Because neurons with large somata have been identified anatomically and functionally in other species (chiefly Hirudo and other hirudinids, but also in Haementeria, a glossiphoniid; Kramer and Goldman, 1981; Kramer and Weisblat, 1985), we studied the spatial distribution of different sizes of neurons within Helobdella ganglion, as candidates for future identification. To correct for systematic differences among animals, the raw size data for each animal were normalized to the mean area of the largest 1% of nuclei in that animal (Fig. S3). We then defined four discrete size classes (small, medium, large and largest), based roughly on breaks in the size distribution among different packets (Fig. 7A; Fig. S3).

The distribution among packets of the smallest nuclei (≤0.20 on the normalized size scale) is about the same as for the distribution of neurons overall (cf. Fig. 7B,C, left). In contrast, the three larger size classes show packet-specific distribution patterns (Fig. 7C); the medium size class (>0.20 and <0.60) is under-represented in AVM and maximally represented in PVM; the large size class (≥0.60 and <1.0) is largely confined to the three anterior packets (LAL, AVM and RAL); and the largest nuclei (≥1.0) are mostly located in the LAL and RAL packets. Both the overall distribution of relative sizes and the distribution of the four size classes among packets were similar in the three individual leeches (Fig. S3). In comparing nuclei of different size classes, we note that the histone-labeled chromatin of some of the largest nuclei has the same or higher fluorescence intensity as in much smaller nuclei (data not shown). This observation is consistent with the hypothesis that different nuclear sizes reflect differences in cell ploidy, but differences in the mCherry signal could also reflect differences among cells in the level of Hau-elav4-dependent expression and/or differences in laser illumination during imaging. Thus, measuring differences in ploidy directly is beyond the scope of the present work.

The implementation of genetic and molecular approaches has made it possible to carry out remarkably detailed investigations of intensively studied models such as mouse, zebrafish, Drosophila and Caenorhabditis. Inevitably, the very power and resource-intensity of these approaches have necessarily restricted the phylogenetic breadth of investigations in modern biology. In particular, these models represent just two of the three main branches of bilaterian animal diversity, the super-phyla Deuterostomia and Ecdysozoa.

The third main branch of animal evolution, the super-phylum Spiralia (Marlétaz et al., 2019), is of interest because it includes the plurality of modern animal phyla. This group is severely understudied compared with the other two, however, and the application of genome editing approaches is correspondingly sparse. Given their phylogenetic position, leeches provide a reference for studying how developmental processes have diverged among diverse animals over half a billion years of independent evolution.

At the time of writing, stable transgenesis in Spiralia has been achieved by microinjecting zygotes for a mollusk, Crepidula (using CRISPR/Cas9 to achieve targeted knockins; Perry and Henry, 2015), for a polychaete annelid, Platynereis (using a Tc1/mariner-type transposable element; Backfisch et al., 2013) and for a planarian flatworm, Macrostomum (using circular or linear DNA fragments; Wudarski et al., 2017). Here, we add to the list of organisms in the superphylum Spiralia for which transgenic approaches have been applied successfully, using a Minos-based transposable element to efficiently introduce transgenes into the genome of a glossiphoniid leech, Helobdella austinensis. This is noteworthy because the Minos-based approach is quite efficient, and because Helobdella is already among the most tractable spiralian embryos. For example, whereas work in most other organisms is largely restricted to zygote injections, the Helobdella embryo is amenable to injecting identifiable, lineage-restricted stem cells or other cells to mark specific somatic lineages in the F0 animals, as well as generating stable transgenic lines by injecting germline precursors throughout early development.

A somewhat paradoxical aspect of our results is that while transgenesis is quite efficient, the copy number of insertions appears to be low. In a typical experiment, most of the surviving embryos have incorporated the transgene, and yet the inverse PCR analysis indicates that just one or two copies of the transgene were heritably inserted. One possible explanation for these results is that embryos with higher copy number insertions die during development. Another possibility is that some intrinsic factors might restrict Minos-mediated insertions to certain ‘hotspots’ in the Helobdella genome. At any rate, the fact that we can readily detect single-locus insertions of transgenes driven by the native Hau-elav4 enhancer suggests that it should be feasible to detect reporters knocked into endogenous genes (expressed at comparable levels), for example by CRISPR/Cas9 mutagenesis coupled with homology-directed repair.

Invertebrate species are used to investigate a wide range of neurobiological phenomena because they offer experimental advantages such as large cells (e.g. the squid giant axon and synapse) and/or cellular simplicity (e.g. Caenorhabditis). Invertebrates also provide fascinating examples of how nervous systems have diverged and specialized among diverse kinds of animals (Bullock and Horridge, 1965).

Individually identified neurons have been characterized in animals ranging in complexity from nematodes (e.g. Caenorhabditis elegans) to fish (Sillar, 2009). As an example of relevance here, the segmental ganglia of medicinal leeches (H. medicinalis and H. verbana) have been used extensively to study neural circuitry and the properties of individual neurons (Calabrese et al., 2016; Kristan et al., 2005; Szczupak, 2023). These studies rely in part on the ability to reproducibly identify individual neurons from ganglion to ganglion, animal to animal, and even among different leech species (Nusbaum and Kristan, 1986). Our work provides the groundwork for mapping identified neurons within the midbody segmental ganglia in Helobdella. Such maps will be useful for future neurobiological analyses in this species.

The extent to which such neural stereotypy exists for animals has not been extensively assessed, with few exceptions. For the nematode C. elegans, serial electron micrograph reconstruction revealed precisely 302 neurons in the hermaphrodite, in 118 morphological classes (White et al., 1986). The initial reconstruction was based on an overlapping combination of series from different regions of multiple individuals, but a large body of subsequent work has effectively confirmed those results. Earlier, serial reconstruction had been used to study the structure and projections of the cyclopic compound eye in an isogenic (parthenogenetic) strain of the water flea Daphnia magna (Macagno et al., 1973). These authors found an essentially invariant cellular structure to the eye, with constancy in the overall number of photoreceptors, in their organization into ommatidia and in the multiplicity of the projections of ommatidia to lamina cells in the optic lobe. Variability emerged in the details of branch morphologies and synaptic connections. Taken to the extreme, this previous work suggests that ‘simple’ systems have rigidly fixed neural composition and largely stereotyped wiring of neural circuits.

For the leech nervous system, the existence of numerous individually identified neurons fits well with the deterministic view (Nicholls and Baylor, 1968). For example, a pair of large serotonergic neurons, known as Rz cells, are readily detected by their serotonin content throughout clitellates, i.e. oligochaetes and leeches (Koza et al., 2006); among leeches, the Rz neurons lie routinely in the AVM packet. Two other examples (from among many conserved cellular phenotypes) are the mechanosensory nociceptive (N) and pressure (P) neurons. Their cell bodies are about the same size, but N cells have their cell bodies in LAL and RAL packets, while the P cell bodies are located in LPL and RPL packets. Homologs of these characterized Hirudo neurons have been identified physiologically in other leech species including glossiphoniid species by both physiological and molecular approaches (Heath-Heckman et al., 2021; Kramer and Goldman, 1981).

A precise anatomical analysis revealed variation in the neuronal content of ganglia in several leech species, but only of a minor nature. Computer-assisted analyses of six serially sectioned and 20 whole-mount ganglia from three hirudinid and one glossiphoniid leech species revealed that the neuronal content in ganglion M10 of H. medicinalis ranged from 389 to 398 for an average of 394±5 (mean±s.d.) and a coefficient of variation of 0.01 (Macagno, 1980). Two other hirudinid species, Macrobdella decora and Haemopis marmorata, were similar in range and consistency for ganglion M10. The same study found the same degree of consistency for Haementeria ghilianii, but this glossiphoniid species contained only around 380 neurons in ganglion M10 (Macagno, 1980).

In the present work, we analyzed the number and size of neuronal nuclei (via the proxy of their nuclear chromatin) in segmental ganglia of a fifth leech species. Helobdella austinensis is a glossiphoniiid species like Haementeria only much smaller in body size. Our data suggest that the variability of neuronal architecture in Helobdella is far greater than found in four larger species. We found that H. austinensis midbody segmental ganglia M02–M18 each contain about 320 neurons, significantly fewer than in the other four species. We also observed a greater range and variability in the number of neurons in H. austinensis ganglia, vis-à-vis the other species. For example, our three counts of H. austinensis ganglion M10 range from 310 to 348 with an average of 328±19 neurons, with a coefficient of variation of 0.06, 6-fold greater than that reported in Hirudo (Macagno, 1980). Other ganglia showed even greater ranges (Fig. 5C). No differences between the ganglia in genital segments (M5 and M6) and non-genital segments were observed for Helobdella in contrast to Hirudo and Haementeria, where these segments contain significantly more neurons than the rest of the segments (Macagno, 1980), but small differences could be masked by the large variability.

Here, we used transgenic approaches that are currently limited to Helobdella, using a pan-neuronal enhancer to drive a readily detected fluorescent protein marker, and the data were obtained by confocal microscopy. These methods allowed us to count a more complete sample of ganglia and eliminated the problem of distinguishing between neurons and non-neuronal cell types in the ganglia. Unavoidably, however, the multiplicity of differences between our experiments and previous work (Macagno, 1980) makes it impossible to distinguish the basis of the incongruent results unambiguously. One possibility is that the observed variance arises artifactually from a methodological error on our part, but this seems unlikely given the high signal-to-noise ratio of the technique. Moreover, objects identified as neurons in some sample packets in the present study were validated by a second author. Another possibility is that the observed variability reflects variation in expression driven by the Hau-elav4 enhancer. This seems unlikely for several reasons: one is the reliability of Elav as a pan-neuronal enhancer in other systems, and in our in situ experiments; in addition, we saw no ‘dark holes’ that might correspond to unlabeled nuclei in the packets during imaging, and the histogram of nuclear intensity shows no significant tail at low values.

Apart from the above-listed technical explanations for the observed differences between our results and Macagno's (1980), more biological explanations suggest themselves as well. Perhaps the most intriguing is the possibility of genuine species differences, perhaps reflecting the great size difference between Helobdella and Hirudo (see following section). Another is that our specimens exhibit developmental abnormalities induced by generations of lab culture or by the transgenesis itself. The latter seems unlikely, as we observed neither an increase in embryonic lethality for the transgenic line compared with wild-type animals (that would be evidence of developmental abnormalities) nor any behavior deficits in locomotion, feeding or reproduction (that would be evidence of malformation in the nervous system).

In addition to the differences discussed above, our more extensive dataset confirms previous conclusions that the neuronal content of the six glia packets is also variable. Despite this variability, we observed conservation in the pattern of distribution of the large neurons. In addition to the pair of large serotonergic neurons (Rz cells), that are typically located in the AVM and which are apparently present throughout the Clitellata (leeches and oligochaetes), Helobdella exhibits additional giant cells in the LAL and RAL packets. Based on the similarities between Helobdella and Haementeria, we assume that these giant cells represent a glossiphoniid-specific pattern, and that the >150-fold range of (intensely labeled) chromatin area/volume across Helobdella neurons represents cell-type-specific rounds of post-mitotic DNA amplification (polyploidy). A question for future work will be to determine whether distinct size classes correspond to particular cell types that undergo controlled amounts of post-mitotic DNA amplification as part of their differentiation program.

One biologically based speculation explaining the observed differences (fewer neurons and higher variability) in Helobdella versus previously examined leech species is based on the dramatically smaller body size of Helobdella compared with Haementeria or Hirudo, and on the differences in scaling parameters for various biological functions. We speculate first that the smaller size means that it is not essential to have one pair of each type of neuron in each segment of the nerve cord; i.e. branches from one neuron of a given type in one segment might serve the functions of that same cell type in adjacent segments and/or on the contralateral side of the animal. Previous work has revealed such interactive and competitive interactions among segmentally homologous neurons in Hirudo, Haementeria and Helobdella (Blair et al., 1990; Gao and Macagno, 1987; Kramer et al., 1985; Kramer and Stent, 1985; Kuwada, 1985; Macagno and Stewart, 1987; Martindale and Shankland, 1990; Muller et al., 1987). If we further assume that ‘redundant’ neurons die in a stochastic manner during normal neural maturation, it could account for both the reduced cell number and the increased variability in cell number between Helobdella and Haementeria. Assuming that this stochastic pruning of ganglionic cell numbers occurs only as the postmitotic neurons are extending their processes, a testable prediction of this model is that segmental ganglia in juvenile Helobdella could have larger and less variable numbers of cells, comparable to those in adult Haementeria.

Regardless of the underlying mechanism, the unexpectedly high variability we observed in the neuronal content of segmental ganglia in Helobdella suggests that this relatively simple nervous system is nonetheless capable of extensive neurodevelopmental plasticity. To investigate these issues with greater precision, it should be feasible to extend the experimental approach introduced here to examine specific subpopulations of neurons, using DNA elements capable of driving cell type-specific reporter expression. Thus, by complementing the rich knowledge of the leech nervous system based mainly on Hirudo, Helobdella is poised to become a promising model system to explore the nature of determinism and plasticity in a simple nervous system.

The authors would like to thank Alexander D. Hartenstein (UC Berkeley) for his pilot work leading to the successful implementation of the Minos-based transgenic protocol in Helobdella, Wei-Yu Tao (National Taiwan University) for screening candidate genes for enhancer cloning, and Zhi-Yi Qiu (National Taiwan University) for his efforts in maintaining transgenic lines. C.J.W. and D.A.W. thank the CRL Molecular Imaging Center for access to the confocal microscope, supported by the Gordon and Betty Moore Foundation (RRID:SCR_017852).