Proteomic analysis of FACS-enriched whole nematocysts from the colonial hydrozoan Hydractinia symbiolongicarpus

TL;DR

The study generated a detailed proteome of stinging cells in the hydrozoan Hydractinia, revealing over 8,000 proteins and comparing them across species.

Contribution

A novel Hydractinia nematocyst proteome was generated using FACS, revealing species-specific and shared venom proteins across cnidarians.

Findings

A Hydractinia nematocyst proteome identified 8,470 proteins, with 760 enriched under a stringent FACS strategy.

Comparison with other cnidarian proteomes revealed shared and Hydrozoa-specific protein clusters.

Overlap with scRNA-seq datasets validated nematocyst-specific proteins in developing and mature cells.

Abstract

Cnidarians possess a cell-based venom system in the form of nematocytes or “stinging cells” that are found across various tissues. The scattered distribution of these venom-containing cells makes isolation difficult, particularly for proteomic studies. These challenges can be circumvented in laboratory systems with efficient culturing conditions and robust molecular resources for downstream validation, such as exists for Hydra and Nematostella. The colonial hydrozoan Hydractinia symbiolongicarpus is an established laboratory model and an emerging candidate for functional studies of the venom system. Here, we present a proteome derived from a fluorescence-activated cell sorted cell population of developing and mature nematocytes from an established Hydractinia transgenic line. We detected a total of 8,470 proteins, of which 2,232 could be statistically quantified across two different fluorescence-activated cell sorting gating strategies. We found that 165 proteins were enriched within a more lenient, low-cell bias strategy while 760 proteins were enriched using a more stringent gating strategy with greater predicted cell viability. We compared this dataset to a previously assembled nematocyst-enriched transcriptome, as well as two different single-cell RNA-sequencing datasets for Hydractinia, to validate the enrichment of protein candidates in the nematocyte lineage. Furthermore, we evaluated orthologous clusters shared between our Hydractinia proteome, a well-established nematocyst proteome from Hydra, and nematocyst-enriched proteomes from two other cnidarians. Through these comparisons, we revealed substantial shared clusters across these four cnidarian species as well as multiple hydrozoan-specific clusters. Overall, our proteomic analysis provides an integral, complementary resource to the established molecular and laboratory tools available in Hydractinia, advancing its utility as a functional venom systems model.

Image 1

•Generation of a nematocyst-enriched proteome for Hydractinia symbiolongicarpus, totaling 8470 detected proteins.•Dual FACS strategy resulted in 760 enriched proteins in a ‘stringent’ small gate versus 165 in a ‘lenient’ large gate.•Overlap with two Hydractinia scRNA-seq datasets validates developing and mature nematocyst-specific proteins.•Comparison with three other cnidarian nematocyst proteomes reveals novel shared genes, including Hydrozoa-specific proteins.

Generation of a nematocyst-enriched proteome for Hydractinia symbiolongicarpus, totaling 8470 detected proteins.

Dual FACS strategy resulted in 760 enriched proteins in a ‘stringent’ small gate versus 165 in a ‘lenient’ large gate.

Overlap with two Hydractinia scRNA-seq datasets validates developing and mature nematocyst-specific proteins.

Comparison with three other cnidarian nematocyst proteomes reveals novel shared genes, including Hydrozoa-specific proteins.

Figures

Full text

Introduction

1

The broader field of “venomics,” the study of venom systems through integrated ‘omics’ technologies, has become greatly enriched by the increased availability and accessibility of genomic, transcriptomic, and proteomic resources (von Reumont et al., 2022). This is particularly true for proteo-transcriptomic approaches used for smaller or difficult-to-collect venomous species, such as invertebrates (von Reumont et al., 2014; Walker et al., 2020). While RNA-sequencing (RNA-seq) based approaches in isolation have greatly expanded our knowledge of venom diversity, protein-level data is important to validate the presence of putative toxins or other venom system-related protein products, especially novel venom components. Even in well studied models, proteo-transcriptomics of venoms has shown that there can be significant variation between RNA expression levels of toxin candidates and the abundance of protein products in crude venom, which has important implications for venom function in an ecological context (Sunagar et al., 2016).

The phylum Cnidaria (jellyfish, hydroids, sea anemones, corals) is widely accepted as the earliest diverging venomous animal lineage, displaying extensive ecological and developmental diversity (Daly et al., 2007). Cnidaria is typified by the presence of extrusive, highly elaborate cellular structures called cnidae or cnidocysts, which are housed in specialized cells called cnidocytes (Fautin, 2009). A subset of cnidocysts are venom-deploying nematocysts, housed in nematocytes, which are found ubiquitously across the phylum (Fautin, 2009). Among diverse cnidarian species, there are between 25 and 30 morphologically distinct nematocysts, which can differ in capsule shape and size as well as patterning of spines and barbs of the tubule (Weill, 1934, reviewed in Mariscal, 1974; Östman, 2000). Cnidarian nematocytes represent one of the few de-centralized animal venom systems, as these cells are distributed across the organism to perform different roles such as feeding, predation, defense, and other interspecific and intraspecific interactions (Ashwood et al., 2020). Multiple distinct nematocyst types are also present in different proportions across tissues within an individual, lending to regionally specific toxin expression patterns (Macrander et al., 2015; Ashwood et al., 2021; Little et al., 2020; Klompen et al., 2022). Multiple studies have taken advantage of the constant replenishment of cnidocytes over the lifetime of the animal by using RNA-seq analysis to determine venom composition (e.g. Liu et al., 2015; Macrander et al., 2015, 2016; Huang et al., 2016; Klompen et al., 2020, 2022; Ashwood et al., 2021). While proteo-transcriptomic analyses are becoming more widely reported, there are limitations to proteomic analysis including tissue availability and technical feasibility.

Various technical challenges have historically hindered proteomic studies on cnidarian venoms (Cantoni et al., 2020; Kaposi et al., 2025). This is in part due to their cell-based venom system; each nematocyst contains only picoliters of toxins (Yanagihara and Shohet, 2012), and these toxins can degrade rapidly after extraction (Othman and Burnett, 1990). However, multiple proteomic studies have yielded important insights into venom diversity across Cnidaria, particularly with regards to medically relevant species for humans from the class Scyphozoa (“true jellyfish”) (e.g. Ponce et al., 2016; Li et al., 2018, 2020; Riyas et al., 2021; Morabito et al., 2015) and class Cubozoa (box jellyfish) (e.g. Carrette and Seymour, 2004; Yanagihara and Shohet, 2012; Brinkman et al., 2015, Cantoni et al., 2020). Venom isolation has also been successful in various sea anemones of the class Anthozoa (e.g. Moran et al., 2013) and members of the class Hydrozoa (e.g. Balasubramanian et al., 2012; Weston et al., 2013; Tassara et al., 2024), though variation in isolation methods renders comparisons between datasets difficult.

Several challenges in nematocyst isolation and venom acquisition have been circumvented using laboratory cnidarian models. Even for relatively small species, there is a high level of tissue accessibility through rigorous culturing conditions, often from clonal laboratory lines, and established models have a growing wealth of molecular and genomic resources. Historically, the freshwater hydroid Hydra vulgaris (Pallas, 1766) (previously H. magnipapillata) was proposed as a chemical biology model (Sher and Zlotkin, 2009). A proteome from isolated Hydra nematocysts remains one of the most well characterized datasets for these cell structures (Balasubramanian et al., 2012). Among the anthozoans (corals and sea anemones), the starlet sea anemone Nematostella vectensis (Stephenson, 1935) is a modern venom system model with multiple studies spanning genomics, transcriptomics and proteomics as well as molecular and functional experiments (e.g. Moran et al., 2013; Sunagar et al., 2018; Sachkova et al., 2020a, Sachkova et al., 2020b; Fridrich et al., 2023; Surm et al., 2024; Kozlovski et al., 2025).

Hydractinia symbiolongicarpus Buss and Yund, 1989 is a colonial hydrozoan used as a laboratory model for developmental and evolutionary biology (Frank et al., 2020). Hydractinia colonies display a division of labor across morphologically distinct polyp types, including gastrozooids that are specialized for feeding and gonozooids that specialize in reproduction (Fig. 1A) (Sanders et al., 2014). These polyps display distinct nematocyst types, including desmonemes and small euryteles in gastrozooids and large euryteles in gonozooids (Klompen et al., 2022). For the purposes of this study, we will primarily be using the term nematocyte and nematocyst, as the other cnidocyte types (e.g. spirocytes) are not present in Hydractinia. This species is highly amenable to genomic manipulation techniques, including transgenesis (Chrysostomou et al., 2022). Klompen et al. (2022) established a nematocyst-specific reporter line in Hydractinia using the promoter for a nematocyst structural protein minicollagen-1 (HsymNcol-1) to drive the expression of the fluorescent marker mScarlet-I (Fig. 1). Minicollagens (abbreviated Ncols) are important structural proteins that form the capsular, and sometimes tubular, scaffolding of the cnida organelle during assembly, making these proteins highly unique to Cnidaria (David et al., 2008). Their distinctive and relatively high expression often lends to their use as a robust marker for the developing cnidocyte lineage, including HsymNcol-1 in Hydractinia (e.g. Klompen et al., 2023). This Hydractinia reporter line was used to isolate nematocytes (mScarlet^+^ cells) through fluorescence-activated cell sorting (FACS) for bulk RNA-sequencing to produce a nematocyst-enriched transcriptome assembly. Based on a customized annotation pipeline, this assembly was used to define a putative venom profile for this species, such as pore-forming jellyfish toxins whose expression was later validated through in situ hybridization and immunohistochemistry (Klompen et al., 2023).Fig. 1Developing and mature nematocy****st-specific transgenic line (HsymNcol-1::mScar***)* of the colonial hydrozoan Hydractinia.** A) Hydractinia colony growing on the edge of a glass slide, showing a mix of gastrozooids or “feeding polyps" (black) and gonozooids or “reproductive polyps” (gray). Polyp silhouettes downloaded and modified from PhyloPic. B) Live confocal image of whole HsymNcol-1::mScar gastrozooid. ∗ indicates oral opening. C) High magnification image from tentacle of HsymNcol-1::mScar gastrozooid, showing mScarlet-I^+^ signal in mature stinging cells mounted across the tentacle, with sensory cnidocil projections visible. Inset shows higher magnification image. D) High magnification image of body column of transgenic gastrozooid, showing mScarlet-I^+^ signal in developing nematocytes where developing capsules are visible. Inset shows higher magnification image. Oval and ellipsoid capsule shapes represent the two main nematocyst types found in gastrozooids. Scale bars: B) 100 μm, C, D) 20 μm, inset: 5 μm.Fig. 1

Here, we employed a dual FACS gating strategy to isolate two mScarlet-I^+^ cell populations for proteomic analysis using liquid chromatography tandem mass spectrometry (LC-MS/MS). We produced a whole nematocyst-enriched proteome and identified over 8,000 proteins using two different annotated genome assemblies. We show that the two gating strategies resulted in differently enriched proteins for each sample. We compared the proteome datasets with two different whole tissue single-cell RNA-sequencing datasets from Hydractina to validate the overlap of numerous developing and mature nematocyte protein candidates. Finally, we performed an orthology analysis of our proteome with three other cnidarian proteomes to better characterize the expression patterns of putative venom system-related gene products across Cnidaria. Together, these approaches have generated a high-quality, protein-level resource for the Hydractinia venom system using a FACS-based collection strategy and illuminated numerous candidate proteins for functional and comparative analyses.

Methods

2

Animal care and imaging

2.1

Animal maintenance and construction of the Hydractinia minicollagen-1:mScarletI (HsymNcol-1::mScar) reporter line are reported in Klompen et al. (2022). Briefly, the transgenic line was produced though random integration of a TOPO pCR2.1 plasmid containing the putative promoter of the native HsymNcol-1, sequence for *mScarlet-*I, and putative terminator of HsymNcol-1 (see Klompen et al., 2022). The plasmid was injected into newly fertilized embryos that were allowed to develop under standard conditions, then progeny screened for mScarlet-I expression as a F0 generation. A single F0 female colony was used to produce F1 progeny, of which a robust F1 female was selected to be maintained for downstream analysis. For live imaging, gastrozooids and gonozooids were isolated from the transgenic colony, relaxed using a 4% MgCl•6H_2_0 solution in 1:1 mix of 35 ppt artificial sea water and DI water, and acquired with a spinning disc Andor DragonFly 200 Confocal Microscope (Oxford Instruments). Images were processed using Fiji (ImageJ2, v2.14.0/1.54f, build: c89e8500e4) (Schindelin et al., 2012; Rueden et al., 2017). All images and figures were produced using Inkscape (v1.4.2, ebf0e940, 2025-05-08) (Inkscape Project, 2025).

Dissociation and FACS

2.2

HsymNcol-1::mScar colonies were starved for 3 days prior to sorting. Approximately 110 polyps (80 gastrozooids + 30 gonozooids) were removed from the colony using a razor blade and relaxed in menthol for 30 min prior to dissociation. Polyps were transferred to a 1.7 mL tube and cleaned twice with calcium and magnesium-free artificial seawater (CMASW). Animals were dissociated in 400 μl of 1% pronase (Sigma-Aldrich, Cat. #10165921001) in CMASW over 1.5 h at room temperature (RT) with constant rocking, as well as gentle flicking of the tube every 15 min. Cells were filtered through a 70 μm Flowmii cell strainer (Millipore Sigma, Cat. BAH136800070) and pelleted at 300×g for 5 min, at 4 °C, prior to sorting. The pellet was resuspended in 1 mL ice-cold 1% BSA in CMASW.

Prior to sorting, DAPI (4′,6-diamidino-2-phenylindole) (1:1000, 1 μl/mL stock) and DRAQ5 (1:200, 5 mM stock) were added to the cell solution. FACS was conducted with the Stowers Institute using an S6 cell sorter (BD Biosciences, USA) with a 130 μm nozzle and 1x sheath (PBS). The sorter was equipped with the 355, 561, and 640 nm laser lines to detect DAPI, mScarlet-I, and DRAQ5, respectively. The resulting cell suspension was evaluated based on side scatter (FSC-A vs. SSC-A), followed by number of singlets (FSC-A vs. FSC-H), mScarlet^+^ cells (mScarlet vs. SSC-A), and by nucleated versus viable cells (DRAQ5 vs. DAPI). Both DAPI and DRAQ5 are used as live-dead indicators in standard FACS experiments; DAPI stains the AT-rich region of DNA with a preference for dead cells where the cell membrane is less viable, while DRAQ5 preferentially stains the nuclei of live cells. For this experiment, both dyes were used as mature nematocyst capsules may also be stained with high levels of DAPI (Szczepanek et al., 2002), and thus appear as "dead cells" in a standard live-dead gating strategy. Two different gating strategies were used for the DRAQ5 vs. DAPI: a large gate (LG) derived from a more lenient strategy to decrease the bias for specific cell types (i.e. maintain cells containing mature nematocysts), and a more conservative small gate (SG) regime to increase cell viability through reduction of “dead” (high DAPI) cells at the potential cost of increased cell bias. In both gating strategies, DRAQ5 staining was maintained at the same “live cell” (i.e. nucleated) cutoff. Cells were kept at 4 °C for the duration of the sort and approximately 300,000 cells for each gating strategy were sorted into respective 1.5 mL tubes coated with CMASW, pelleted briefly, flash frozen and stored in −70 °C until further use. Sorting was evaluated using FACSDiva v9.1.2 (BD Biosciences, USA).

Sorts consisting of DAPI-only and DRAQ5-only stained cells from wildtype colonies as well as an unstained transgenic (mScarlet-I only) sample were performed as gating controls before the experiment. Representative images of the gating strategy for the experimental sort as well as gating controls are presented in Fig. 2 and Suppl. Figs. S2–S5. Prior sorting experiments validated that developing and mature nematocytes were enriched in sorted samples using microscopy. FSC files for the sorting experiment in this study are available from the Stowers Original Data Repository (see Data Availability).Fig. 2Proteins identified from LC-MS/MS analysis of FACS-enriched nematocyte samples using two sorting strategies. A) Representative gating from FACS experiment of HsymNcol-1::mScar polyps, where gating for mScarlet^+^ cells is shown above and two different gating strategies across DRAQ5^+^ versus DAPI^−^ cells are shown below, namely a small gate (SG; red) and a large gate (LG; purple). Of note, the LG is inclusive of the SG strategy but expands the DAPI threshold to include additional DAPI^+^ cells (y-axis). B) Number of proteins identified from FAC-sorted cell samples and subsequent LC-MS/MS analysis, including enriched cells in the small (red) and large (purple) gate samples. C) Volcano plot indicating the differentially expressed protein abundances across quantified proteins enriched in the SG and LG datasets. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)Fig. 2

Protein extraction

2.3

Sorted cells were lysed through freeze/thaw/sonicate cycles. The LG and SG cells were kept on ice and submerged 1 min in liquid nitrogen (LN_2_). The cells were thawed at RT, vortexed, and incubated in LN_2_ twice more prior to sonication. A 10-min sonication cycle comprising 10 rounds of 30 s ON/30 s OFF was performed on the samples kept in an ice/water bath. The tubes were centrifuged at 14,000×g for 10 min, and the supernatant was dried using a SpeedVac concentrator (Savant) without heat. The dried protein pellets were stored at −20 °C.

Protein digestion and high pH reverse phase fractionation

2.4

Proteins were solubilized in 30 μl of 8M urea, 100 mM Tris, pH8.5, vortexed, and sonicated 10 times with 30 s ON/30 s OFF cycles in a water bath at RT. Proteins were reduced by adding tris(2-carboxyethyl)phosphine (TCEP at 1M) to 5 mM final at RT for 30 min. Reduced cysteine residues were carboxymethylated by adding 5 μl of 2-chloroacetamide (CAM, made fresh at 0.5 M) and samples were incubated for 30 min in the dark at RT. Endoproteinase Lys-C at 0.1 μg/μl was added at 1:1000 w/w and the digestion proceeded for over 6 h at 37 °C. The samples were diluted to 2M urea with 100 mM Tris, pH8.5, 2 mM CaCl_2_, and sequencing grade modified trypsin (Promega, #V511A) was added at 1:200 w/w and incubated at 37 °C overnight. The digested protein samples were centrifuged at 16000×g for 30 min and transferred to new tubes. A fluorometric peptide assay (Pierce, #23290) was performed on 10 μl of each digested sample according to manufacturer's instructions. Peptide amounts were ca. 400 ng and 200 ng for the LG and SG samples, respectively. Peptides were dried using a SpeedVac concentrator.

The dried peptide mixture was resuspended in 300 μl of 0.1% trifluoroacetic acid (TFA) and loaded onto one high pH fractionation cartridge (Pierce, cat. # 84868) placed on a new 2 mL sample tube. After centrifuging at 3000×g for 2 min, the eluate was collected as the “flow-through” fraction. The loaded cartridge was placed on a new 2 mL sample tube, washed with 300 μl of ddH20, and the elution collected as “wash” fraction. A total of 8 HpH RP fractions were collected by sequential elution in new sample tubes using 300 μl of 10%, 12.5%, 15%, 17.5%, 20%, 22.5%, 25%, and 50% acetonitrile in 0.1% TFA. The solvents were evaporated to dryness using vacuum centrifugation.

Data acquisition from liquid chromatography tandem mass spectrometry (LC-MS/MS)

2.5

Peptides were analyzed on an Orbitrap Eclipse Tribrid Mass Spectrometer (Thermo Scientific) with a FAIMS Pro interface, equipped with a Nanospray Flex Ion Source, and coupled to a Dionex UltiMate 3000 RSCLnano System (U3000). A 75 μm i.d. analytical microcapillary column was packed in-house with 250 mm of 1.9 μm ReproSil-Pur C18-AQ resin (Dr. Masch). An AgileSLEEVE (Analytical Sales & Products) was used to maintain column temperature at 50 °C. The analytical column was equilibrated and conditioned twice by loading 50 ng of peptides digested from Saccharomyces cerevisiae whole cell extract. Indexed Retention Time peptides (iRT kit, Biognosys) were loaded and analyzed twice to ensure adequate chromatographic separation and a total ion intensity above 10e^9^.

Peptides from each of the 8 HpH RP fractions were solubilized in 22 μl of buffer A (5% acetonitrile in 0.1% formic acid, FA) and loaded via a 20 μL sample loop on an Acclaim PepMap 100 C18 trap cartridge (0.3 mm inner diameter (i.d.), 5 mm length; Thermo Fisher Scientific) with the U3000 loading pump at 2 μL/min via the autosampler. The organic solvent solutions were water:acetonitrile:formic acid at 95:5:0.1 (volume ratio) for buffer A (pH 2.6) and 20:80:0.1 (volume ratio) for buffer B. Peptides were eluted directly into the mass spectrometer over a 95-min chromatography gradient with a nano pump flow rate set to 0.180 μl/min. The Orbitrap Eclipse was set up to run the MS OT/ddMS2 IT HCD method with two FAIMS compensation voltages (CV) at −40V and −60V and a cycle time of 1 s. Briefly, eluting peptide ions were scanned from 375 to 1500 m/z in the Orbitrap at 120,000 resolving power before data dependent (dd) MS2 fragmentation by HCD at 35% Normalized Collision Energy (NCE) and detection in the ion trap set to rapid scan rate. The isolation window was 1.6 m/z and dynamic exclusion, with low and high mass tolerances set at 10 ppm, was enabled for 45 s.

Data processing and analysis from LC-MS/MS

2.6

Two databases of Hydractinia symbiolongicarpus protein sequences were downloaded from NCBI RefSeq and NHGRI (Hydractinia Genome Project Portal, 2025) in June 2025 (see Data Availability). The NCBI RefSeq and NHGRI FASTA files, as well as a file containing 470 common contaminants were concatenated, and redundant sequences were removed. In cases where the NCBI RefSeq and NHGRI databases contained representatives of the same sequence, preference was given to the NCBI RefSeq entries. This resulted in a total of 40,301 non-redundant (NR) sequences, including 24,732 entries from NCBI (“XP_” accessions), 15,099 entries from NHGRI (“HyS” accessions), and 470 contaminants (“Cont_” accessions). This database was imported to Proteome Discoverer 3.4 (PD3.4). Hsym_LG-HpH and Hsym_SG-HpH raw files were opened in PD3.4 and a search was set up using default parameters for the “Processing” and “Consensus” workflows, except “Fixed Value PSM Validator” node was used as a less stringent way to filter the peptide spectrum matches by FDR when the overall protein abundance in the samples is very low.

The “proteins” output was exported from PD3.4 (Supp. Table S1**, Sheet 1**) and the sequences were further filtered by removing contaminants, removing proteins detected by single peptides, and only considering proteins with label-free quantitative values (signal intensities from mass spectrometry spectra to determine the relative protein or peptide abundance) and adjusted p-values (Supp. Table S1**, Sheet 2**). PD3.4 calculated adjusted p-values for single-replicate samples using a background-based t-test, which assessed the significance of a protein change against the background distribution of all protein/peptide ratios in the dataset. It also replaced 0 values (i.e. not detected/quantified in one sample) by a low abundance value to calculate Fold-Changes (FC), i.e. FC = 0.01 and FC = 100 when not detected in numerator and denominator, respectively. Proteins with LOG_2_Fold-Change(LG/SG) <-2 or >2 and -LOG_10_(adj. p-value) >1.35 were considered enriched in the SG and LG fractions, respectively (Fig. 2C–Supp. Table S1**, Sheet 3**). The volcano plot and subsequent pie charts were produced using the tidyverse package (Wickham et al., 2019) in R (v4.5.0) (R Core Team, 2025) through RStudio (v2025.5.0.496) (Posit Team, 2025) adjusted using Inkscape.

Data annotation and comparative analysis

2.7

FASTA files from associated identified proteins were annotated through the eggNOG-mapper (v2.1.12) through the online web browser (Cantalapiedra et al., 2021) using the eggNOG 5 resource (Huerta-Cepas et al., 2019), and the associated PFAM domain matches as well as eggNOG descriptions were isolated. The total protein FASTA was additionally queried using DIAMOND blastp (v2.1.13.167) (Buchfink et al., 2021) with the -sensitive flag against UniProt/SwissProt (The UniProt Consortium, 2025) and custom databases produced from the Hydractinia nematocyst-enriched RNA-seq assembly and venom profile produced in Klompen et al. (2022) (“TRINITY_” accessions; NCBI TSA: GJUZ00000000.1). Subsequent annotations as well as a summary of proteomic analysis outcomes, including quantifiable proteins and those enriched in either gating strategy, are available in Suppl. Table S2**, Sheet 1.**

Two different single-cell RNA-sequencing (scRNA-seq) datasets were used to determine overlapping protein sequences across the nematocyte lineage (Salamanca-Díaz et al., 2025; Song et al., 2025) (For the purpose of version control, data collection was originally performed using the (Song et al., 2025) preprint deposited in bioRxiv on June 2025). Marker genes in clusters that were specifically annotated as either developing or mature nematocytes/cnidocytes were collected and used to compare overlapping genes using InteractiVenn (Heberle et al., 2015) (Fig. 3A–C). All sequence headers were converted to RefSeq gene identifiers (i.e. “LOC”) using a conversion table produced from a DIAMOND blastp search of CDS sequences downloaded from the NHGRI assembly and queried against the protein models for the RefSeq assembly using the -max-target-seqs 1 flag. Representative gene expression patterns from each scRNA-seq cell atlas were downloaded from respective publicly available browsers (Fig. 3D, see Data Availability). A summary of annotated clusters used in these comparisons can be found in Suppl. Table S2**, Sheet 2**.Fig. 3Overlap of Hydractinia whole nematocy****st proteome with developing and mature nematocyte clusters derived from Hydractinia scRNA-seq. A-C shows Venn diagrams of overlapping genes from Hydractinia whole nematocyst proteome dataset versus marker genes in associated developing and mature nematocyte clusters in both Salamanca-Díaz et al. (2025)and Song et al. (2025) (seeSuppl. Table S2**, Sheet 2** for cluster annotations). Shapes in each Venn diagram correspond to the presence of genes outlined in D. All Venn diagrams are modified from InteractiVenn (Heberle et al., 2015). A) All detected Hydractinia proteins against all cluster markers. Numbers in bold below are the number of genes that were statistically quantifiable in the Hydractinia proteomic dataset. B) Hydractinia proteins enriched in either the LG or SG proteomics dataset compared to scRNA-seq clusters specific to developing nematocytes (nematoblasts or cnidoblasts). C) Hydractinia proteins in either the LG or SG proteomics dataset compared to scRNA-seq clusters specific to mature nematocytes (or cnidocytes). D) Specific clusters from each scRNA-seq dataset (Suppl. Table S2**, Sheet 2**). Three representative genes and expression patterns, along with putative annotations (Suppl. Table S2**, Sheet 1)**, including a LG-enriched, mature nematocyte marker (triangle), LG-enriched, developing nematocyte marker (star), and putative MACPF toxin identified in Klompen et al. (2022) that is SG-enriched in developing nematocytes (hexagon). Note that the MACPF toxin appears to be expressed in a single specific cluster in each scRNA-seq dataset. Representative images were downloaded and modified from respective cell atlas browser webpages (see Data Availability).Fig. 3

For orthology analysis, a FASTA file containing all identified proteins from nematocyst-enriched tissue samples in Hydra (Balasubramanian et al., 2012), Sanderia malayensis, and Rhopilema esculatum (Leung et al., 2020) were input into OrthoVenn3 server (Sun et al., 2023) using the OrthoML algorithm, E-value cutoff at 1e^-5^, inflation value at 1.5, and annotation and protein similarity enabled, as well as phylogenetic analysis enabled using the WAG + CAT model. Clusters associated with shared orthologous genes across all four species (“shared”; Suppl. Table S2**, Sheet 3**) or those shared only between Hydra and Hydractinia (“Hydrozoa-specific”; Suppl. Table S2**, Sheet 4**) were compared with annotation categories from the Hydra nematocyst proteome (Balasubramanian et al., 2012) (Fig. 4).Fig. 4Orthologous clusters derived from four medusozoan nematocyst****-enriched proteomes using OrthoVenn3. A cladogram representing phylogenetic relationship of the four species used in this work is shown that corresponds to the following datasets: Hydractinia whole nematocyst proteome (blue), isolated nematocyst proteins from Hydra (purple), Sanderia malayensis (green), and Rhopilema esculatem (orange). UpSet Plot is modified from the output of the OrthoVenn3 browser. Clusters shared between all species (“shared”) and those shared only between Hydra and Hydractinia (“Hydrozoa-specific”) are shown in black and outlined in gray boxes. Representative pie charts from the “shared” and “Hydrozoa-specific” groups indicate annotation categories from representative Hydra sequences (Balasubramanian et al., 2012) in each associated cluster (Suppl. Table S2**, Sheet 3 and Sheet 4)**. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)Fig. 4

Results

3

Dual sorting strategies produced distinct proteomic profiles

3.1

In total, we identified 8,470 total proteins across the stringent, small gate (SG) and more lenient, large gate (LG) sorting-strategy samples using a non-redundant protein annotation dataset from two genome assemblies in Hydractinia, NCBI RefSeq (Kon-Nanjo et al., 2023) and NHGRI (Schnitzler et al., 2024). Of these, 4,438 were detected only by a single peptide, though these included several putative toxins and nematocyst-specific structural genes. Label-free quantitative values and adjusted p-values were determined for 2,232 proteins. Among these, 760 proteins were enriched in the SG sample and 165 were enriched in the LG sample, respectively. Within the SG sample, 697 of these 760 enriched proteins were unique to the SG sample. In the LG sample, 117 of the 165 enriched proteins were found to be unique (Fig. 2). A summary of these findings for each detected protein as well as broad annotations based on the EGGnog wrapper and best matches against the UniProt SwissProt database are available in Supplemental Table S2.

We next sought to determine how our proteome compared to a previously assembled nematocyst-enriched transcriptome that used the same reporter line (Klompen et al., 2022). First, the transcriptome assembly was downloaded from NCBI (TSA: GJUZ00000000.1). Putative coding regions were predicted using TransDecoder (v5.5.0), and redundancy was removed using CDHit (v4.8.1) (Fu et al., 2012). Of the 8,470 proteins we detected in this study, 7,937 (93.7%) matched a predicted transcript, with 7,650 (90.3%) having an E-value less than 1e^-10^, suggesting strong correlation between the two datasets. Conversely, 66 of 105 venom-like genes annotated from the transcriptome assembly were detected in the proteome. Notably, however, 34/66 of these proteins were detected by a single peptide, and only 22/66 were quantifiable (Suppl. Table S2, Sheet 1). Of these 22 quantified putative venom proteins, eight were not only enriched in the SG strategy but were unique to the SG sample. Only two putative toxins were enriched in the LG sample, but neither of these two putative toxins were unique to the LG sample.

Comparison of Hydractinia whole nematocyst proteome versus multi-omics hydractinia datasets

3.2

For both the NCBI RefSeq and NHGRI assemblies, multi-tissue Hydractinia scRNA-seq cell browsers are publicly available (Schnitzler et al., 2024; Salamanca-Díaz et al., 2025; Song et al., 2025; see Data Availability). We took advantage of these datasets to compare our proteome with marker genes for clusters annotated as developing nematocytes (nematoblasts) and mature nematocytes. In order to determine if either gating strategy enriched for specifically the developing or mature developmental stages, we evaluated the overlap of marker genes in each of the nematocyte-associated clusters within the scRNA-seq datasets against proteins in our proteome. This included a broader overview of the total proteins detected in our proteome across all marker genes of the nematocyte-associated clusters as well as the overlap of marker genes specific to developing or mature cnidocyte-specific clusters against proteins enriched in the LG or SG datasets (Fig. 3). In order to make these comparisons, sequences across all datasets (i.e. scRNA-seq from Song et al. (2025) and our proteome, as needed) were converted to gene identifiers from the NCBI RefSeq genome (“LOC”). Notably, the RefSeq genome contains more splice-aware genes than the NHGRI genome, thus conversion of NHGRI sequence identifiers can result in multiple matching LOC identifiers. In our study, the 8,470 detected proteins in our dataset when converted resulted in 8,179 total associated “LOC” identifiers for comparison.

Of the total detected proteins in our Hydractinia proteome, 733 (9% of the proteins detected) were also identified as marker genes of developing and mature nematocytes within the two scRNA-seq datasets, including 373 proteins that were identified within all three datasets. Of note, marker gene tables are limited in that they do not represent the total expression of genes in specific clusters and typically are used to detect genes highly specific to a single cluster. Thus, genes that might be specific to multiple clusters (i.e. multiple developing nematocyte clusters) in a single atlas are not necessarily detected in a marker table analysis. Marker genes in the two scRNA-seq studies were also generated using different statistical methods, which can result in variable markers even from the same data (Salamanca-Díaz et al., 2025). This suggests there may be some nematocyte-specific genes that are not identified in the marker tables and thus missing from these comparisons. Future scRNA-seq analyses and/or bulk RNA-seq of isolated nematocyte populations are required to identify these additional putative proteins. Altogether, of the 2,446 proteins that were quantifiable in our study, 300 (12% of quantifiable proteins) overlapped across the two scRNA-seq datasets. Of these, 162/300 overlapped with both datasets, 76 only with the Salamanca-Díaz et al. (2025) dataset, and 62 only with Song et al. (2025) (parenthesis, Fig. 3A). These 162 quantifiable proteins in our proteome and identified as marker genes in both scRNA-seq datasets represent highly conserved candidate proteins for future study. However, these comparisons also show there are prominent differences between each of the scRNA-seq experiments. This could be due to biological variation of the colonies used in each experiment or technical differences in sample collection, preparation, and downstream analysis. Both considerations warrant additional validation experiments.

We further compared proteins that were enriched in the SG and LG sorted samples and marker genes specifically associated with either developing nematocytes (Fig. 3B) or mature nematocytes (Fig. 3C). Our hypothesis was that the LG strategy includes a higher proportion of cells containing mature cnidae, as it encompasses the SG with the additional leniency in the DAPI signal, which would favor matured cnida capsules that are DAPI^+^ (Szczepanek et al., 2002). In the developing nematocyte clusters in both scRNA-seq datasets, there was greater overlap of SG-enriched proteins, including a putative MACPF-toxin also identified in the venom profile from the bulk nematocyst-enriched RNA-seq (TRINITY_DN5188_c0_g1_i3; LOC130649358, HyS0006.76) (hexagon, Fig. 3D–Suppl. Table S2**, Sheet 1**). Expression patterns in both cell atlases suggests that this putative toxin is highly enriched in a specific developing nematocyst cluster, implying a degree of cell type specificity. While this may suggest the SG sample is enriched for developing nematocytes in comparison to mature nematocytes, the greater number of overlapping proteins could also be due to the higher relative number of enriched genes overall in the SG sample compared to the LG sample. This is supported by the presence of two different genes within the LG-enriched dataset that are identified either within the developing (LOC130612262, HyS0052.52) or mature (LOC130614102, HyS0002.36) markers, validated by the associated expression patterns in each cell atlas (triangle, star, Fig. 3D). In both the developing and mature nematocyte comparisons against the LG- or SG-enriched proteins, we found no genes overlapped with all four datasets.

Comparison of the Hydractinia nematocyst proteome with other cnidarian nematocyst proteomes

3.3

To identify orthologous proteins that may be present across other cnidarian groups, we compared our results to a well-characterized Hydra (Hydrozoa) proteomic dataset from whole isolated nematocysts (Balasubramanian et al., 2012) as well as two jellyfish tentacle proteomes produced from Sanderia malayensis and Rhopilema esculateum (Scyphozoa) (Leung et al., 2020). Using OrthoVenn3 (Sun et al., 2023), we found that 77 orthologous clusters were shared between all four cnidarian species while 39 clusters were shared only between Hydractinia and Hydra (Fig. 4). Using representative candidates from the Hydra datasets in each cluster, we evaluated different annotation categories within both cluster groups (Fig. 4) (Suppl. Table S2**, Sheet 3 and 4)**. While structural proteins, venom, and various enzymes were a high proportion in both groups, a higher relative number of novel proteins (those without prior annotations) were found in the Hydrozoa-specific clusters (13/39) versus the shared clusters (6/77), suggesting that these novel proteins may be important nematocyst-related genes that are class-specific.

Discussion

4

Production of a Hydractinia whole nematocyst proteome

4.1

Our whole nematocyst proteome is derived from FACS-enriched developing and mature nematocysts from a transgenic Hydractinia colony, in which a similar protocol was previously used to produce a nematocyst-enriched RNA-seq experiment (Klompen et al., 2022). Historically, isolation strategies for cnidarian species focus on enriching intact nematocyst capsules, often using density gradient-based approaches such as Percoll or 1 M citrate (Weber et al., 1987; Yanagihara and Shohet, 2012). The major limitation to these studies is that the cellular material from mature nematocytes (or more broadly cnidocytes) are likely removed as the intact capsule dislodges from the cell. Our study takes advantage of a reporter line for developing and mature nematocytes, enabling the generation of a comprehensive nematocyst-enriched proteome across cell development. Thus, our proteome is a valuable resource for evaluating both venom toxin candidates as well as putative protein products involved in the assembly of nematocysts (e.g. structural proteins), as shown by the overlap of both developing and mature markers in both previously reported scRNA-seq datasets and our dataset.

Of note, our proteome was developed using a small quantity of tissue. Approximately 110 polyps (∼1 mm in size/polyp) were used for the collection of ∼600,000 cells (across two samples), including ∼80 gastrozooids (Fig. 1) and ∼30 female gonozooids (Suppl. Fig. S1). A higher proportion of gastrozooids were used with the intention of increasing the number of developing nematocytes from the gastrozooid body column, as this is a known active site for nematogenesis (Fig. 1). While gonozooids typically display little active nematogenesis compared to the gastrozooids, the mature nematocytes in the capitate tentacles of gonozooids typically are mScarlet^+^, though at reduced levels compared to actively developing nematocytes (Suppl. Fig. S1). Thus, we expect our proteome to contain representative proteins from all types of nematocytes across development in gastrozooids and gonoozooids. Importantly, each polyp type displays distinct nematocyst types, namely small euryteles and desmonemes in gastrozooids and large euryteles in gonozooids (Klompen et al., 2022). Large and small euryteles appear morphologically similar, and thus may have similar structural proteins, but there is evidence to suggest they contain different toxins (Klompen et al., 2023). ScRNA-seq is beginning to predict distinct molecular profiles of nematocyst types across Hydractinia (Salamanca-Díaz et al., 2025; Schnitzler et al., 2025; Song et al., 2025), and our proteome will be a valuable complementary tool for exploring candidate proteins within this cell lineage. Furthermore, given the relatively little tissue required for our FACS-based strategy, this study shows a proof-of-principle for future studies in Hydractinia and other cnidarians where transgenesis is feasible (i.e. Nematostella, see Kozlovski et al., 2025).

Evaluating protein enrichment from two FAC-enriched nematocyte populations

4.2

In this work, we employed a dual FACS-enrichment strategy using a more “stringent” SG strategy and “lenient” LG strategy. We show that this dual gating strategy had a significant influence on the resulting proteome, where the more stringent SG strategy resulted in a higher number of identified proteins, though at the potential cost of increased cell bias (i.e. reduction of mature nematocytes). It remains unclear exactly why the more restrictive SG strategy resulted in higher numbers of identified proteins; presumably the higher cell viability would benefit downstream analysis, but it is also possible the SG strategy enriched for cell types or cell states more well suited to proteomic methods in this study. However, this work does showcase how the selection of gating strategy is critical for experimental repeatability, in particular for this cnidarian cell type. Such variation in gating strategies may result in incorrect interpretations of protein expression patterns, particularly if the protein abundance is already at low levels. In the case of this experiment, multiple previously predicted toxin proteins identified from a nematocyst-enriched transcriptome by Klompen et al. (2022) were enriched or specific to the SG sample, which emphasized greater cell viability. A more lenient LG-like regime may be preferred to increase the number of cells for downstream analysis, but based on the results of this work, it may also lead to the incomplete conclusion that some toxin peptides are not present in a sample.

The increased proportion of toxin-like genes may imply that the SG gate dataset is enriched specifically for developing nematocytes, since venom production occurs prior to the maturation phase during nematocyst development. This is further emphasized by the higher overlap of developing nematocyte markers within the SG-enriched proteins. For example, a putative MACPF-like toxin protein (LOC130649358, HyS0006.76) that was also identified in the nematocyst-enriched transcriptome assembly (TRINITY_DN5188_c0_g1_i3) is specific to developing nematocyst clusters in both scRNA-seq datasets, as well as a unique protein identified in the SG sample. Interestingly, this toxin is specific to a singular cluster of developing nematocytes in both atlases, suggesting that these represent a specific subpopulation of nematocytes (Fig. 3D). This potentially suggests this toxin is specific to a single morphological type, such as the small euryteles of gastrozooids. This type of expression pattern has previously been shown for a jellyfish toxin venom protein, HsymJFT1c-I (LOC130642394, HyS0055.81), in the small euryteles of gastrozooids (Klompen et al., 2023).

However, the higher overlap of the SG proteins with the developing (and mature) nematocyte scRNA-seq datasets may be because of the higher number of enriched proteins overall within the SG dataset compared to the LG proteins. This could be explained by higher viability overall of cells sorted from the SG, and not enrichment of a specific cell state (e.g. developing nematocytes over mature nematocytes). This is further evident given that both developing and mature genes are found in the LG-enriched dataset. For example, an LG-enriched gene (LOC130614102, HyS0002.36) annotated as a glutathione synthetase is found to be specific to the mature clusters in scRNA-seq. It has been shown that reduced glutathione activates the feeding response in Hydra (Loomis, 1955) and triggers cnidocytes in sea anemones (La Spada et al., 2002) through a still undetermined receptor (though see Gavazzi et al., 2023). This implicates the identified glutathione synthetase is active during preparation for nematocyst maturation and eventual discharge. But another LG-enriched gene (LOC130612262, HyS0052.52) annotated as a sulfatase is specifically found within the developing nematocyte clusters, which may represent an important and specific enzyme in the assembly of the nematocyst structure. Due to the lack of replication between these two gating strategies in the current experiment, future downstream analyses such as localization-based experiments (e.g. immunolocalization) or formation of transgenic lines will be needed to validate these putative protein expression patterns across nematocyte development.

Implications from orthology predications across multiple cnidarian nematocyst proteomes

4.3

Our expectation was that shared orthologous proteins from nematocyst-enriched proteomes from other cnidarian species would elucidate potential structural proteins and enzymes that are important to the formation of nematocysts. We found multiple previously identified structural genes specific to cnidarian nematocytes, including those categorized as minicollagen-like (Kurz et al., 1991; Klompen et al., 2022; reviewed in David et al., 2008), nematocyst outer wall antigen (NOWA) (Engel et al., 2002), and cnidarian proline-rich protein (CPP) (Balasubramanian et al., 2012). There were also several candidate calcium modulators, which are known to be important to regulate extracellular calcium necessary for nematocyst discharge (Lubbock et al., 1981; Thorington and Hessinger, 2023), as well as multiple enzymes, peptidases, and venom toxins. Only 6/77 clusters corresponded to novel proteins, of which four were considered not to be cnidarian specific (Balasubramanian et al., 2012) (Suppl. Table S2, Sheet 3).

This contrasts with the clusters specific to Hydra and Hydractinia where 13 of the 33 total clusters are annotated as novel proteins, of which only five were categorized as “not exclusive to Cnidaria” (Suppl. Table S2, Sheet 4). This implies that these clusters may represent Hydrozoa-specific proteins in the nematocyst structure and/or involvement in nematocyst assembly. Hydrozoa displays the greatest variations in nematocyst morphologies, with an estimated 17 subtypes being exclusive to the group (Mariscal, 1974), thus these novel proteins are interesting candidates for further functional analysis. These Hydrozoa clusters also include known structural proteins, including minicollagen-15 and spinalin that have been well-characterized in specific nematocyst structures in Hydra (Koch et al., 1997; Adamczyk et al., 2008), implying these may also be present in Hydractinia nematocysts. Multiple enzymes, peptidases, and putative venom toxins are also identified in this group, thoough at lower proportions than the “shared” cluster profile.

While this study did not characterize all putative clusters in depth, we observed interesting patterns across the dataset. For example, the cluster with the highest count number (1,077) was shared between the Hydractinia, Sanderia, and Rhopilema samples. It is possible this is a result of the collection method in each study as well as the methedology of the proteomic analysis. For the Sanderia and Rhopilema samples, whole tentacles were used for nematocyst extractions and submitted to nano-flow LC-MS/MS using tentacle transcriptomes as the search database, resulting in over 3,000 proteins detected from each species (Leung et al., 2020). By contrast, the Hydra proteome was produced from highly concentrated nematocyst samples from whole specimens that were solubilized and run through a one-dimensional gel, of which each gel slice was submitted for mass spectrometry. This resulted in approximately 400 total proteins detected (Balasubramanian et al., 2012). Given over 8,000 proteins were detected in our Hydractinia dataset, the strong overlap with Sanderia and Rhopilema is likely due to the increased sensitivity of the methodology in the MS/MS experiments. The second highest cluster count (864) was between Sanderia and Rhopilema, which likely is due to their close phylogenetic relationship as well as similarity in collection method and analysis (Leung et al., 2020). The third highest cluster (733) is composed of Hydractinia singlets, which implies that our collection strategy (FACS-enrichment) was more sensitive than either of the three other samples.

Hydractinia as a model for functional and comparative venomics

4.4

Hydractinia, as well as the other cnidarian model systems like Clytia, Hydra and Nematostella, is an exception to a current lack of functional laboratory models in venomous animal studies. These cnidarian models have well established culturing methods and closed life cycles, meaning tissue is accessible across the year for experiments. Furthermore, there is an opportunity to utilize modern ‘omics’ resources in venom-focused studies, including chromosome-level genome assemblies, robust annotated gene models, scRNA-seq, Assay for Transposase-accessible chromatin with sequencing (ATAC-seq), and spatial technologies. Cell-specific methods such as scRNA-seq are particularly useful as cnidocytes are often observed as isolated clusters due to high expression levels of cnidocyte-specific genes. Hydractinia already has many of these resources publicly available as well as validated functional genomics techniques such as transgenesis (e.g. DuBuc et al., 2020; Klompen et al., 2022), gene knockdown (e.g. Quiroga-Artigas et al., 2020), gene knockout (e.g. DuBuc et al., 2020) and knockin (e.g. Sanders et al., 2018), and several studies have already targeted questions related to the venom system in this species and others (Fig. 5). For example, a similar FACS-based isolation approach using cnidocyte-specific Nematostella transgenics have provided valuable insights into the gene expression and regulation of developing cnidocytes (Sunagar et al., 2018; Kozlovski et al., 2025). With the addition of the whole nematocyst-enriched proteome generated by this study, Hydractinia is well equipped as a functional venom system model, in particular within a comparative context of other cnidarian models such as Hydra and Nematostella (Fig. 5).Fig. 5Summary of molecular resources and tools available to study the venom system of Hydractinia, with representative studies. Blue indicates currently available “omics” datasets and example publications. Yellow indicates functional genomics techniques that can be utilized for future venom system studies with example publications. Citations in bold are directly focused on the study of the venom system of Hydractinia. Purple indicates the potential of the Hydractinia system to be used in comparative ‘venomics’ studies with the historical venom model Hydra and modern venom model Nematostella. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)Fig. 5

Conclusion

5

In this work we present a nematocyst-enriched proteome for the colonial hydrozoan Hydractinia using a FACS strategy to collect whole nematocyte (or cnidocyte) cell populations from an established transgenic line. To our knowledge, this is the first report of a proteomic dataset for nematocysts using this isolation strategy. By using FACS-enrichment of specific cell populations, it is possible to reliably and repeatedly collect similar cell populations across multiple experiments with relatively little tissue. For Hydractinia, this strategy can be coupled with genomic manipulation experiments against candidate toxins to provide empirical functional understanding, complemented by publicly available genomic and molecular datasets in this system. Through comparative analysis of two Hydractinia single-cell RNA-seq datasets, as well as three other cnidarian nematocyst proteomes, we identified numerous venom system-related candidates across the development of nematocytes at the transcriptomic and proteomic level. Many of these candidates are currently uncharacterized and putatively cnidarian, or even hydrozoan, specific. The availability of this high quality whole nematocyst proteome will have a meaningful impact on future studies both within the broader cnidarian community and for venom research focused on the evolution, ecology, and diversity of animal venom systems.

CRediT authorship contribution statement

Anna M.L. Klompen: Writing – review & editing, Writing – original draft, Visualization, Methodology, Funding acquisition, Formal analysis, Data curation, Conceptualization. Kevin Ferro: Writing – review & editing, Methodology, Formal analysis. Cassandra G. Kempf: Writing – review & editing, Methodology, Formal analysis. Laurence Florens: Writing – review & editing, Formal analysis, Data curation. Matthew C. Gibson: Writing – review & editing, Supervision, Funding acquisition. Paulyn Cartwright: Writing – review & editing, Supervision, Funding acquisition, Conceptualization.

Ethical statement

The animals used in this study do not require NIH Institutional Animal Care and Use Committee approval for research purposes. None of the animals used in this study are endangered species.

Funding

This research was supported by funding through the 10.13039/100007795Stowers Institute for Medical Research as well as the 10.13039/100000001National Science Foundation (DEB-2153774, awarded to PC). AMLK further acknowledges support from the KU Chancellor's Doctoral Fellowship as well as 10.13039/100000001National Science Foundation Postdoctoral Research Fellowship in Biology (DBI-2208988) during portions of the project duration.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.