Gene Expression Underlying Teloblast Formation in Leech (Annelida: Hirudinae)

Abstract

Objective:

Species of glossiphoniid leech provide model systems for exploring fundamental questions in evolutionary and developmental biology. In this study, we took advantage of large embryonic cells and stereotypical cleavage patterns in the leech, Helobdella austinensis, to identify genes associated with the births of teloblasts and their lineage specifications.

Methods:

Using staged embryos and pools of dissected precursor cells and teloblasts, a systematic computational comparison of staged and cell type-specific transcriptomic data was employed to identify unique gene sets associated with mesodermal (M), neuroectodermal (N) and teloblast (M + N) cell formation.

Results and Conclusions:

Our predicted candidate genes comprised sets of nucleic acid binding factors as anticipated but displayed little similarity with differentiation factors in other metazoans (eg, mammals, cnidarians), suggesting that stem cell/lineage induction processes have diverged across the Animalia and may reflect fundamentally different modes of embryonic development.

Author Summary

Leeches are historically well known for bloodletting practices that were widespread throughout the 19th century and were also a classic preparation for neurobiologists during the same period. In modern times they have found medical applications in reconstructive surgeries, wound healing and more recently “hirudotherapy.” Lesser known are the contributions of leeches to embryology and evolution. Here we take advantage of unusually large embryos in the leech, Helobdella autinensis, to isolate homogeneous populations of teloblasts and their immediate precursor cells. By extracting RNA from these different cells, we were able to identify sets of genes that were unique to each cell type. Computational analyses of targeted genes suggest that many are associated with determining stem cell fate and lineage specification (eg, mesoderm, neuroectoderm). Interestingly, however, our predicted gene sets display only marginal similarity with comparable genes in other animals (eg, humans, Hydra), implying that early developmental processes associated with stem cell formation have diverged considerably over the course of animal evolution.

Introduction

Leeches comprise a major clade within the superphylum, Lophotrochozoa, a taxonomic lineage separated from more commonly studied model species within Ecdycozoa (eg, Drosophila melanogaster, Caenorhabditis elegans) and Deuterostomia (eg, Mus musculus, Xenopus laevis). Species of glossiphoniid leeches display ovipary and many have large, yolky embryos that develop rapidly, making them suitable model systems for studying developmental processes in a representative spiralian.^1,2 In particular, early embryonic cleavages are asymmetric and lead to the genesis of bilateral pairs of embryonic stem cells (M, N, O/P, and Q teloblasts) that give rise to segmental mesodermal and ectodermal tissue. ³ Mesodermal teloblasts arise from the symmetric cleavage of precursor cell DM, while ectodermal teloblasts arise from asymmetric cleavages of the NOPQ lineages, respectively (Figure 1).

Figure 1.

Early developmental cleavages in Helobdella leech embryos. Overt asymmetry appears upon first cleavage (stage 2) and thereafter cell lineages display stereotypical cleavage patterns leading to precursors DM and DNOPQ (Stage 4b) and teloblasts M, N, O/P and Q (stages 4c-6b). Mesodermal teloblasts (M) arise from the symmetric cleavage of precursor cell DM (stages 4b, 4c), while ectodermal teloblasts arise from asymmetric cleavages of the NOPQ lineages (stages 5-6b), respectively. Shaded areas identify teloplasm. Derived from Weisblat and Kuo. ⁴

One realization from comparative embryology over the past few decades has been the striking similarity of developmental processes across metazoan phyla, as well as regulatory genes associated with birth defects and diseases.^5,6 These observations suggest that fundamental aspects of developmental genetic circuitry have been conserved among Animalia since the Cambrian radiation, ~545 million years ago. By this reasoning, one might hypothesize that genes associated with stem cell formation and function are conserved, for example between annelids and other bilaterian phyla, including chordates (eg, human).

Stem cell specification is a topic of considerable interest in contemporary biology due to potential applications in tissue replacement, regeneration and 3D organ printing.^7-9 A logistic breakthrough was achieved by the identification of 4 transcription factors—Oct3/4, Sox2, Klf4, c-Myc—whose overexpression appears to induce stem cell pluripotency in diverse mammalian somatic cells, ¹⁰ and which reflects endogenous embryonic stem cell (ESC) differentiation in mouse.^11,12 These methodologies have led to avenues for generating human stem cells ex vivo, thus circumventing previous requirements for extracting equivalent cells from human embryos or tissue. ¹³

Interestingly, the aforementioned Yamanaka factors do not align well with stem cell-associated genes in non-bilaterians. For example, the Cnidarian polyp Hydra appears to employ different genetic modules for maintaining 4 distinct stem cell lineages, characterized by almost unlimited regenerative capacity.^14-16 Although stem cell molecular signatures in Hydra include overlapping functional motifs (eg, Zinc finger, Leu zipper), a different complement of factors specify stemness (ie, FoxO, TCF, ZNF28, HES1b, Ski), suggesting that the mammalian ESC core transcriptional network may be a later invention in bilaterian evolution. ¹⁴

To determine the extent to which stem cell formation is conserved across the Metazoa, we profiled Chordata and Cnidarian stem cell-associated homologs in developing leech embryos to decipher whether their regulated expression was consistent with orthologous function(s) across these disparate animal phyla. Our data imply that the genetic network underlying “stemness” appears to be markedly different among metazoans, which may reflect fundamentally different modes of embryonic development.

Results

To identify putative overlapping pathways associated with stem cell induction and maintenance across metazoan phyla, we used canonical mammalian (ie, Oct3/4, Sox2, Klf4, c-Myc, Nanog) and Hydra (ie, FoxO, TCF, ZNF28, HES1b, Ski) stemness factors to identify homologs in leech Helobdella reference genome NCBI GCF_000326865.2. BLASTP was conducted using conserved domains, respectively, and hits exceeding an ~70% motif conservation threshold were selected as likely homologs. Corresponding genes were profiled across early developmental stages in H. austinensis transcriptomes, as well as putative stem cell precursors (D, DM, DNOPQ, NOPQ) and teloblasts M and N. By these criteria, mammalian and Hydra homologs in leech were not differentially expressed during leech embryogenesis; rather, most appeared at low to moderate levels during early, pre-teloblast stages (1 -4) and thereafter fluctuated only slightly in staged embryos and cell types associated with teloblast birth (Figure 2). Likewise, genes including piwi, pumilio, and vasa that have been linked with regeneration and adult stem cell activation in annelids,^17,18 were not differentially regulated in our dataset (Supplemental Figure 1).

Figure 2.

Gene expression profiles of canonical stemness induction/differentiation factors during leech embryogenesis. Top panel shows leech homologs of Yamanaka factors associated with ESC specification in mammals, ¹⁰ lower panel shows the equivalent in Hydra. ¹⁴ Staged embryos and dissected cell types are shown on X-axes, each spot represents an independent transcriptome. Predicted homologs are indicated by colored spots, respectively. Y-axes (ie, normalized transcript #’s) are scaled differently to better resolve data points in each respective panel.

This apparent divergence between mammalian, Hydra and leech stem cell regulatory networks prompted us to conduct an unbiased search for genes differentially expressed at the junction between precursor cells/teloblasts in H. austinensis. To conduct these analyses, we compared gene expression profiles among staged embryos and relevant cell types using conventional approaches (ie, DESeq2 ¹⁹ ), as well as absolute- and fold-changes (ie, highest numerical differences and highest values of fold-change, respectively), and transcripts with no representation (zero) in one of the comparative cells (Figure 3). Factoring the developmental decision in question, we gauged that relatively small percentage changes observed in DESeq2- and absolute-identified transcripts were not necessarily critical differentiation factors, even though relative differential-transcript numbers were high. Thus, we also considered fold-changes (including from zero) and generated separate groups of upregulated/downregulated transcripts associated with teloblast formation (ie, occurring in both M and N) and lineage specification (ie, occurring in M or N) (Supplmental Table 1). Following the exclusion of predicted housekeeping genes (eg, ribosomal proteins, transporters, proteases), transcripts without Helrodraft matches (eg, possible contamination) and/or those with DeepFri scores below 0.3, our combined data from all comparative approaches yielded sets of genes encoding nucleic acid binding proteins we propose most likely to influence teloblast cell fate (Supplmental Table 2). The same approach was applied within the 2 selected teloblast lineages, M and N, to identify lineage-specific transcripts (Supplemental Tables 3 and 4). With the exception of nanos homolog Hro-nos, ²⁰ a downregulated transcript during early development in H. robusta and detected by our analysis (see Supplmental Table 1), most developmental regulators characterized in leech to date were not represented in our precursor→teloblast gene dataset, including leech homologs of engrailed, ²¹ hairy and Enhancer of split, ²² even-skipped, ²³ bmp, ²⁴ twist,^25,26 hedgehog, ²⁷ wnt, ²⁸ dorsal and snail, ²⁹ and notch ³⁰ (Supplemental Figure 2).

Figure 3.

Unbiased searches of differentially-expressed gene transcripts during teloblast formation in leech. (A) DM versus M. (B) DNOPQ versus N. (C) DNOPQ versus N. (D) NOPQ versus N. Black spots forming the diagonal comprise non-differentially expressed transcripts in Helobdella austinensis. Four independent approaches were employed to identify differentially expressed transcripts between comparative cells: DESeq2 ¹⁹ (red), Absolute (blue; highest numerical differences), Fold (green; highest values of fold-change) and Zero (yellow; with one comparative cell not expressing the transcript). Up-regulated transcripts appear above the diagonal and down-regulated are below.

Interestingly, our comparisons between precursor cells and differentiated teloblasts revealed that cell NOPQ—previously assumed to be a teloblast precursor cell ³¹ —displayed patterns more consistent with teloblast gene expression. Specifically, teloblast-specific transcripts co-expressed in both M and N with respect to their closest common parental cell D aligned more closely with DNOPQ versus NOPQ comparisons than with NOPQ versus N (Figure 4), as supported by coefficients of variance (Figure 5). These data imply that stemness in the DNOPQ lineage is acquired upon the birth of NOPQ and that subsequent divisions specify lineage fate restrictions (ie, N, O/P, and Q).

Figure 4.

Double stage-transition expression profiles support stemness for cell types M and NOPQ. (A) D→M versus D→N. (B) DM→M versus DNOPQ→N. (C) DM→M versus DNOPQ→NOPQ. (D) DM→M versus NOPQ→N. Putative stem-related transcripts appear in upper right quadrant as green spots (upregulated) and lower left quadrant as red spots (downregulated). Teloblast M more closely aligns with expression in NOPQ than N (compare panels (C) and (D)).

Figure 5.

Correlation coefficients of stage transition expression profiles. In the context of overlapping gene expression at major stage transitions (see Figure 4), green bars indicate a stage transition compared to itself (normalized to 1.0), blue bars show positive correlations, red bars show negative correlations. Cell type NOPQ consistently displayed a “stemness” pattern as compared to birth of the N teloblast.

Discussion

To date, ESC-related studies have focused on mammalian systems which narrow key stemness genes down to a handful of canonical transcription factors necessary for induction and maintenance.^10,32 These genes, however, do not appear to be reflective of stem cell genesis in other animal phyla (eg, Cnidaria) or even other Bilateria, based on the current study. Specifically, our unbiased search for candidate genes associated with teloblast formation in leech yielded a unique set of differentially expressed transcripts encoding a mixture of uncharacterized nucleic acid binding proteins. Without functional data for these candidate genes, we cannot determine the core genetic requirements for establishing stemness in leech; nonetheless, it is noteworthy that minimal overlap in stem cell-associated transcriptional machinery is observed between disparate animal groups ranging from Hydra to mammals and leech. Of relevance is the function of c-Myc, which comprises one of the 4 Yamanaka factors necessary to confer stemness in mammalian cells, ¹⁰ but which appears to be absent altogether in the genomes of Hydra and leech, respectively ¹⁴ (see Figure 4). It remains unclear whether this disparity represents an evolutionary novelty in Mammalia and/or whether this more closely aligns Cnidaria (a basal metazoan) with the Annelida, in the context of stem cell genesis.

The composition of stem-related genes identified in our study shows that no obvious homologs were apparent in comparison with other known stem cell regulators. But note that common stemness genes are not well documented among metazoan phyla and it could be that animal lineages diverged genetically from a common archetype but mechanistically are using core (but different) sets of transcription factors to trigger similar downstream processes (eg, pluripotency, self-renewal). Alternatively, the fundamentally different modes of development observed between metazoans may have selected specialized genetic circuitry to guide very different embryological processes. Mammals, for example, generate pluripotent ES cells within the inner cell mass of the nascent blastula, which then give rise to all 3 germ layers during gastrulation. In contrast, early cell divisions in leech generate lineage-restricted teloblasts giving rise to mesodermal (M) and ectodermal (N, O/P, Q) tissue, while endoderm forms from blastomeres A, B, and C (see Figure 1). From this perspective, and factoring differences in stem cell potency (ie, pluripotency in mammals vs unipotency/multipotency in leech), it is perhaps not surprising that minimal overlap in stem cell gene expression was observed between these disparate taxa. Note also that our criterion for stem-associated transcripts in leech (ie, expressed in both M and N teloblasts) could also select for other teloblast functions (eg, asymmetric cell cleavages).

Clearly, our study represents only a first step in identifying putative genes associated with stemness in annelids and will require the functional dissection of these genes/encoded proteins in developing embryos. Nevertheless, our data analysis permitted the identification of lineage-specific factors differentially expressed in mesodermal (M) and neuroectodermal (N) lineages, respectively. These data suggest that lineage-specificity is determined by a small number of transcriptional regulators having minimal sequence similarity with known metazoan factors (see Supplemental Table 1), including those previously identified in early leech development (see Figure 4). Taken together, these observations suggest that conventional candidate gene approaches may not reveal some core developmental mechanisms in outlier systems (eg, leech), and that cell fate/lineage decisions may not be exclusively under transcriptional control but rather could be influenced by other mechanisms (eg, post-transcriptional, translational).^33-35

The detailed profiling of early developmental cells in this study also revealed an important temporal decision in the context of acquiring a stem-like genetic profile. Embryonic development in leech is characterized by stereotyped cleavages that occur within a well-defined timeline.^1,36 In the D macromere, which gives rise to bilateral pairs of stem cell teloblasts M, N, O/P, and Q, an oblique cleavage of D yields a smaller DM and larger DNOPQ cell, respectively, at stage 4b (see Figure 1). Following a few rounds of cell divisions that generate micromeres (2 in DM, 3 in DNOPQ), cells DM and DNOPQ divide equally within a narrow time window at stage 4c to produce M_L/M_R and NOPQ_L/NOPQ_R, respectively. According to our data, this developmental time point represents the junction between precursor cells and stem cells, namely the birth of cells M_L/M_R and NOPQ_L/NOPQ_R, even though the latter will thereafter subdivide into formally recognized teloblasts N, O/P, and Q. This developmental strategy may minimize the complexity of coordinating stemness and lineage-restricted cell fate in the DNOPQ lineage, that is, stem cell induction need only occur once (DNOPQ → NOPQ_L/NOPQ_R) as opposed to its iteration across multiple cell cycles with variable numbers of micromeres in between teloblast births (ie, N, O/P, Q). From a conceptual standpoint, more straightforward genetic circuitry may reduce the possibility of developmental anomalies, and could also explain apparent discrepancies with previous attempts to identify stem cell-associated factors in leech.^31,37 In those studies, NOPQ was designated as the non-stem cell precursor to N, and thus differentially expressed transcripts would not include stem cell induction factors since both cells had already acquired stemness.

Conclusion

Collectively, our computational analyses of staged and cell type-specific gene expression profiles has generated a set of gene candidates that may be associated with stem cell teloblast induction and lineage specification in leech. Their functional characterization and identification of downstream targets are necessary to gauge the extent to which underlying genetic circuitry has diverged among bilaterians (eg, mammals) and more distant metazoans (eg, Hydra).

Materials and Methods

Specimens and Cell Acquisition

A colony of glossiphoniid leech, Helobdella austinensis, was derived from a founder population originally collected in Austin, TX.^38,39 Leeches were maintained in glass bowls at 23°C and fed a diet of freshwater snails and insect larvae, as described.^26,40 Staged embryos were harvested from representative clutches corresponding to designated developmental time points (see Figure 1). Between 100 and 200 embryos were collected at each stage, typically from 2 to 3 independent clutches (each containing ~30-80 embryos). Specific cell types, namely D, DM, M, DNOPQ, NOPQ, and N, were dissected with fine pins from fixed embryos [~20 minutes in an ACME solution containing acetic acid:methanol (11:1)] at corresponding stages, respectively. Only a single cell was extracted from each embryo, remaining cellular debris were discarded. At least 100 cells of each type were pooled and immediately transferred into a guanidium solution (ie, lysis buffer; Total RNA Extraction Kit, ZYMO Research), incubated for ~1 hour at room temperature to solubilize tissue and stabilize RNA, and then frozen at −80°C for storage. Upon the accumulation of sufficient cell numbers, total RNA was extracted according to the manufacturer’s instructions (ZYMO Research) and stored at −80°C.

Transcriptome Construction

RNAs from staged embryos (at least 3 independent replicates at representative stages containing >100 embryos per stage) were shipped on dry ice to Azenta (Plainfield, NJ), which processed samples for RNA-seq using the Helobdella robusta reference genome (NCBI taxis: 6412; http://genome.jgi-psf.org/Helro1/Helro1.home.html) for annotation. Transcriptomes included more than 20 million reads, with >15 000 unique transcripts aligning to the H. robusta genome. ⁴¹ Staged transcriptomes were normalized by Azenta as part of their bioinformatics package using DESeq2, ¹⁹ which scales by a sample-specific factor corresponding to total gene hit counts (Supplemental Figure 3). Cell type-specific RNAs from pooled cells of D, DM, DNOPQ, NOPQ, M, and N were processed at the Single Cell and Transcriptomics Core Facility of the Johns Hopkins School of Medicine (Baltimore, MD). Libraries were prepared using Illumina’s Stranded mRNA Prep Ligation kit (Illumina, Inc., San Diego, CA) with 50 ng total RNA input. RNA concentrations were measured with Nanodrop and RNA quality was checked on Agilent Fragment Analyzer using High Sensitivity total RNA assays. Library concentrations were measured with Qubit and quality was checked on Agilent Fragment Analyzer using the High Sensitivity Large Fragment method. Libraries were normalized to 2 nM to prepare a library pool, and concentrations were verified by Fragment Analyzer and qPCR assays. Library sequencing of 4 technical replicates from each sample, from single end, was performed on a NextSeq High Output 75 cycle flow cell, to generate FASTQ data files.

Comparative Computational Analyses

Transcripts were trimmed by Trimmomatic v.0.36 ⁴² prior to alignment with Helobdella robusta reference genome NCBI (GCF_000326865.2), then aligned using STAR. ⁴³ Hence, all H. austinensis genes utilized a H. robusta proxy name. Statistical analysis and graphing were executed using Python3, along with NumPy and SciPy libraries. ⁴⁴ In comparisons of stage transitions, a factor 0.1 was added to expression of all stages to avoid division by zeroes. A “Fold-Zero” ranking method was developed by taking the top 20 candidates identified by fold and zero ranking lists, respectively, combining the lists and then sorting by relative changes in transcript level. Likewise, ranking within the “Absolute” category was based on highest→lowest numerical differences between transcript levels. DESeq2 ¹⁹ was conducted in the R package to identify differentially expressed genes with Wald’s test of significance at a threshold of alpha-value > .05, then genes were ranked using adjusted P-values (Padj). Note that ranking provided a methodical way to list transcripts across the different comparative approaches but did not influence the selection of putative genes associated with teloblast formation (eg, predicted nucleic acid binding proteins) Graphs were generated with Matlab and the “Matplotlib” API for Python. For plots with total transcript counts, average values from the various trials were recorded. Scatter plots were generated using average counts with Log 10 applied to both axes. For co-expression between M and N lineages, a 10-fold threshold for each stage transition was used above/below the value at which genes were expressed at that transition. Stemness genes comprise transcripts with expression in both stage transitions of interest (ie, M and N) and ordered according to the sum of fold differences.

Functional Prediction Methods

Prosite ⁴⁵ and associated results were obtained through the Web interface (https://prosite.expasy.org/) for the domain group search and results were written into csv format in Python. NCBI BLAST in all versions were used from its web implementation (https://blast.ncbi.nlm.nih.gov/Blast.cgi). We used the core nucleotide database for general searches, and “Nucleotide Collection” for biased “a priori” searches. Neural network methods DeepFri were implemented and executed with scaling 0-1 according to their Github documentation ⁴⁶ (https://github.com/flatironinstitute/DeepFRI). Implementation of DeepGoPlus ⁴⁷ used the training data set “data-1.0.12” from May 18, 2023.