Zebrafish EEC transcriptional signatures identify EEC progenitors and six distinct EEC subtypes

To understand EEC subtype diversity and differentiation in zebrafish, we used established markers of secretory cells [9,28,29,34–47] to identify and subset these populations from published intestinal scRNA-seq datasets derived from larval [48] and adult [49] zebrafish (S1A and S1B Fig). Integration of these secretory cells from larval and adult datasets showed the expected structure of a hub of secretory progenitor cells with spokes of distinct secretory populations (S1C and S1D Fig). With this confirmation, we then subsetted and reintegrated just EECs and secretory progenitors into a new combined larval and adult dataset (Fig 1A and 1B). Within this dataset, we identified clusters 0–6 as EECs based on neurod1 expression (Fig 1C) and clusters 7–10 as secretory progenitors (S2A Fig). We found that each EEC cluster contained both larval and adult cells with no apparent age dependency for any EEC subtype (Figs 1D, S2B, and S2C), suggesting that EEC subtypes and their underlying gene expression programs are surprisingly consistent between larval and mature adult stages in zebrafish.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. scRNA-seq of larval and adult EECs identifies 7 distinct clusters.

(A) Schematic overview of identifying secretory cells from published larval [48] and adult [49] datasets. (B) UMAP of secretory progenitors and EECs from larval and adult zebrafish identifies 11 clusters. (C) Normalized expression of the zebrafish pan-EEC transcription factor and reporter neurod1 labels clusters 0–6. (D) The percentage of cells derived from either the larval (blue) or adult (green) datasets in each of the 7 EEC clusters. (E) Scaled expression dotplot of relevant marker genes of the 7 EEC clusters. Dots represent gene expression in the specified cluster. Dot radius corresponds to the percent of cells expressing the gene and dot transparency corresponds to expression intensity. Expression is reported for both the larval (blue) and adult (green) cells in the cluster, and genes not annotated in a given dataset are noted with an X. EEC hormone coding genes are marked with an asterisk next to their gene name on the x axis. Black asterisks indicate genes we identified peptides from and gray asterisks mark those with no detected peptides (Figs 2B and S4). Note, annotations of gast (see scRNA-seq methods), tac4b (see scRNA-seq methods, S15 Fig), and sct (see Results and Figs 2 and S6) were not previously identified and were defined in this paper. (F) Venn diagram of EEC hormone coding gene homologs found in zebrafish and mammals. (G) Heatmap of expression of transcription factors significantly enriched in each EEC cluster in both larval and adult cells. (H) UMAP colored by pseudotime as determined by Monocle 3 when cluster 7 was provided as the origin. Brighter coloring indicates “older” cell age, and the gray line is the predicted differentiation trajectory. Underlying data can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003522.g001

To identify the shared features of larval and adult EECs, we assessed the expression of EEC markers and hormone coding genes in the combined dataset (Fig 1E). Hormone coding genes (marked with an asterisk in Fig 1E) are often used to separate EECs into subtypes. In accord, we found that expression of each EEC hormone coding gene is largely partitioned into a single cluster with remarkable consistency across larval and adult stages. All neurod1+ clusters expressed EEC hormone coding genes except for cluster 0, which instead expressed neurog3. This pattern is consistent with features of EEC progenitors in mammals, which are marked by Neurog3 expression and have not yet begun to express hormone coding genes [10–14,50]. As such, we operationally defined cluster 0 as a putative EEC progenitor population and clusters 1–6 as mature EEC subtypes. We observed substantial overlap between the EEC hormone coding genes expressed in our zebrafish dataset and those reported in published mammalian datasets (Fig 1F) [19,23,24,51–55], highlighting the critical role of these signaling molecules during vertebrate evolution.

To test if the EEC subtype populations we identified are consistent across studies, we examined the EECs in another published larval zebrafish scRNA-seq dataset (S3A Fig) [56]. After selecting and re-clustering the EECs in that additional dataset, we found that the EEC hormone genes we highlighted in Fig 1E partitioned into clusters remarkably similarly to what we observed in our dataset (S3B Fig). The one exception was that ghrl and pyyb were expressed in two separate clusters in our combined dataset, but one in the additional dataset. However, the expression of ghrl and pyyb was non-overlapping in the additional dataset with ghrl primarily expressed at the boundary between the pyyb+ cluster and a cluster enriched for secretory progenitor markers (S3C Fig). This indicates that the zebrafish EEC subtypes described here are relatively consistent across independent studies.

To explore the broader transcriptional programs driving EEC subtype identity, we identified enriched genes specific to each cluster (S1 Table, S2D Fig). Using existing ontologies [57], we identified transcription factors that were significantly enriched in both larval and adult cells in each cluster (Fig 1G). We also performed pseudotime analysis [58], a computational method for ordering cells along a hypothetical timeline based on transcriptomic similarity that enables the study of cell differentiation dynamics. Using the secretory progenitor cluster 7 as the start point, this analysis predicted a differentiation trajectory from secretory progenitor clusters to EEC subtype clusters (Fig 1H). These complementary approaches highlighted certain patterns in EEC populations. For example, we note that putative EEC progenitor cluster 0 is enriched in ascl1a and sox4b in addition to neurog3. While mammalian Neurog3 is the most canonical EEC progenitor marker, studies have also shown that Ascl1/ascl1a and Sox4/sox4b homologs label endocrine progenitor populations in mammals [19,24,41,59–61] and zebrafish [28,31,62,63]. EEC subtype cluster 1 is enriched for insm1a and nkx2.2a whose mammalian homologs Insm1 and Nkx2.2 are known to play a role in the differentiation of Neurog3+ EEC progenitors into mature EEC subtypes [64–67]. Together with the pseudotime data, these findings support a classification of cluster 0 as EEC progenitors and subtype cluster 1 as potentially an early EEC state. Interestingly, subtype cluster 1 also expresses the EEC hormone coding genes ghrelin (ghrl), somatostatin (sst2), motilin-like (mlnl), and gastric inhibitory peptide (gip), which are canonically thought to mark a mature EEC population [50,68]. EEC subtype clusters 2 and 3 appear to have a close relationship given their adjacent positions in pseudotime and the high expression of subtype cluster 2 enriched transcription factors in subtype cluster 3. EEC subtype cluster 3 is uniquely enriched for the known endocrine transcription factor pdx1 [69–71] as well as osr2, which has been shown in mammals and zebrafish to be more highly expressed in the mid and distal small intestine [23,72], indicating a potential regional distribution of this subtype. Finally, subtype cluster 6, which expresses the rate-limiting enzyme in serotonin synthesis tph1b (Fig 1E), is enriched for transcription factors lmx1ba and lmx1al whose orthologs in mammals have been shown to be important for the differentiation of serotonergic neurons and EECs, respectively [66,73,74]. Serotonergic EECs, termed enterochromaffin cells, separate from other EECs in mammalian scRNA-seq studies [19,24,25,75]. Likewise, in our dataset, we find that subtype cluster 6 most clearly separates from the remaining EEC populations (Fig 1B). Altogether, this identifies cluster 6 as zebrafish enterochromaffin cells and suggests the divergence between enterochromaffin cells and other EECs is an ancestral trait to fishes and mammals. This in silico analysis of scRNA-seq data highlights a limited set of transcription factors that might play a role in promoting differentiation from secretory progenitors into discrete EEC subtypes.

Zebrafish EECs produce fully processed and modified peptide hormones in vivo

EEC peptide hormones are derived from the initial translated preprohormone through several steps of specialized cleavage and posttranslational modification (Fig 2A). As a point of nomenclature, in this text we refer to the peptide-coding gene and transcript in italics (GHRELIN for human, Ghrl for mouse, and ghrl for zebrafish and other model organisms), the pro-peptide protein with a leading capitalization (Ghrelin), and the secreted peptide in lowercase (ghrelin). To understand what processed peptides are produced by zebrafish EECs, we performed mass spectrometry on sorted EECs from larval and adult samples. Due to technical limitations, adult samples contained roughly twice as many cells as larval samples, and peptides derived from adult samples are thus overrepresented in our peptidomic dataset (see Methods). In S2 Table, we report each peptide and the quantitation (area under the peak value) from each of the 3 larval and 3 adult biological replicate samples. However, we did not perform comparisons between values from larval versus adult samples as the lower cell content in larval samples meant that low or absent signal in larval samples could be due to the limit of detection of the instrument and not reflective of the biology of the system. Additionally, the EEC peptide data presented in Figs 2 and S4 only reports the detection of unique individual peptides and should not be interpreted as reflective of their natural relative abundance in EEC secretory granules.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Peptidomic analysis of zebrafish EECs reveals production of conserved and novel peptides.

(A) Schematic of peptide hormone processing using the human EEC hormone coding gene GHRELIN as an example. (B) Number of peptides identified from each zebrafish hormone coding gene ranked by normalized mRNA expression level as assessed by scRNA-seq (S1 Table). Note, annotations of gast (see scRNA-seq methods), tac4b (see scRNA-seq methods, S15 Fig), and sct (see Results and Figs 2 and S6) were not previously identified and were defined in this paper. (C) Correlation between the normalized mRNA expression of zebrafish EEC hormone coding genes and the number of peptides detected. (D) Representation of ghrl-derived peptides detected in zebrafish EECs. The primary amino acid sequence is shown at the top in black with observed missense variants indicated in red above. Blue horizontal lines below the amino acid sequence represent the individual peptides detected in our study with small black vertical lines denoting the stop and start of each peptide. Abutting peptides share a single vertical line to represent the stop of the preceding peptide and the start of the subsequent one. Different colored squares represent various posttranslational modifications detected. Yellow shading marks the region aligning to the functional ghrelin peptide in humans and teal shading marks the region aligning to the functional obestatin peptide in humans. (E) Alignment of the primary amino acid sequence of the Ghrelin protein in zebrafish (Uniprot F1QKX9), goldfish (Uniprot Q8AUU1), catfish (Ensembl ENSIPUT00015057246.1), salmon (Uniprot B3IYK1), trout (Uniprot A0A060WHA4), bullfrog (Uniprot Q90W22), chicken (Uniprot Q8AV73), mouse (Uniprot Q9EQX0), rat (Uniprot Q9QYH7), human (Uniprot Q9UBU3). Amino acids are color-coded based on their percent identity match across all the reported species with darker coloring indicating a more conserved residue. Human functional peptide annotations are labeled below, including the conserved acylation at the serine 3 residue. (F) Representation of peptides detected from the zebrafish sct gene. Annotation is as described in D, with pink shading marking the region aligning to the functional secretin peptide in humans and gray shading marking the region highly conserved across fishes. (G) Alignment of the primary amino acid sequence of the Secretin protein in zebrafish (Ensembl ENSDARG00000099233), Sumatra barb (NCBI LOC122330700), grass carp (NCBI LOC127507777), Pachon cavefish (Ensembl ENSAMXG00005000672), fugu (Ensembl enstrug00000026212), razorfish (NCBI CAJ1056004), gar (NCBI LOC107075768), elephant shark (NCBI LOC121850333), mouse (Uniprot Q08535), rat (Uniprot P11384), dog (Uniprot A0A813PAI0), pig (Uniprot P63298), human (Uniprot P09683). Coloring and annotation as described in E. Underlying data can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003522.g002

Of the 23 hormone-coding genes transcriptionally expressed by zebrafish EECs (Fig 1E), we identified peptide fragments aligning to 19 of them (Fig 2B). Genes with no detected peptides included neurotensin (nts), tachykinin precursor 1 (tac1), and tachykinin 4b (tac4b), which were very lowly expressed at the transcript level, and brain-derived neurotrophic factor (bdnf), whose peptide product is too large for optimal detection with our methodology. Of note, while expression of tph1b is used as a proxy for serotonin because it catalyzes the rate-limiting step in its formation, serotonin itself is derived from a single amino acid and is not suited to detection with our methodology. The EEC peptides we identified include products of hormone coding genes specifically expressed by EEC subtypes 1–6, suggesting each of those subtypes produce peptide hormones. We found a modest correlation between the expression level of a gene in scRNA-seq data (S1 Table) and the number of unique peptides detected that align to that gene (Fig 2C). We also identified peptides from other common components of the hormone-containing secretory vesicles released by EECs. Namely, we detected peptides from proprotein convertase subtilisin/kexin type 1 and 2 (pcsk1 and pcsk2), known for their role in specialized cleavage of peptide hormones [76], and secretogranin 2a, 2b, 3, and 5, which help form the secretory vesicles and can be processed into hormones themselves (S4Q–S4X Fig) [77–80]. Together, these observations indicate that our EEC peptidomic approach effectively isolated and defined peptide contents of zebrafish EEC secretory vesicles.

Using this dataset, we generated an atlas of zebrafish EEC peptides including their sequence, posttranslational modifications, and alignment to homologous proteins in humans and other vertebrates. This is provided as a resource in the supplemental materials supporting this article (S4 Fig), and we highlight Ghrelin-derived peptides as an example (Fig 2D). In mammals, two main peptide hormone products from the GHRELIN gene have been described, ghrelin and obestatin [81,82]. Mammalian ghrelin is a 28-amino acid peptide that is often fatty acylated at the serine 3 residue [83]. While both acylated and deacylated ghrelin peptides are produced, fatty acylation improves binding to ghrelin’s receptor and thus is critical to its function [81,83]. As such, the serine 3 residue is highly conserved across species with evidence of acylation at this site in each vertebrate examined (Fig 2E) [81,84,85]. Indeed, studies in goldfish, a close relative of zebrafish, identified a functional 17-amino acid ghrelin peptide that is acylated at the serine 3 residue [84,86], and intraperitoneal injection of acylated (but not desacylated) ghrelin stimulated increased feeding in goldfish and grass carp [86,87]. Mammalian obestatin is a 22-amino acid peptide that was identified more recently by bioinformatic prediction [82], and its function is still debated [82,88]. We detected 52 Ghrelin-derived peptides in zebrafish, including 10 that align with the portion of the protein shown to encode a functional ghrelin peptide in humans and goldfish (Fig 2D and 2E) [86,89,90]. As fatty acylation is key to ghrelin’s function and not included in the posttranslational modification parameters of our initial search, we performed a second search for just ghrelin peptides and included octanoyl and decanoyl modifications in our parameters. Peptides with both octanoyl- and decanoyl-modified serine 3 residues were detected, demonstrating remarkable conservation of the processing and posttranslational modification of this peptide hormone (Figs 2D, S5A, and S5B). While functional studies remain to be performed, the alignment and modification data suggest these peptides could be endogenous zebrafish ghrelin. In contrast, while we detected peptides aligning to portions of the reported obestatin region, we did not detect any peptide spanning that entire region. Our data suggests obestatin may not be produced in fishes, although co-injection of a synthetic predicted obestatin-like peptide has been shown to inhibit the appetite-stimulating effects of injected ghrelin in grass carp [87].

While our data on Ghrelin-derived peptides confirmed our predictions based on studies in closely related organisms, we were also able to leverage our peptidomic dataset for novel discoveries. For example, we detected several peptides aligned to the largely unannotated gene si:zfos-2372e4.1 (Fig 2F). Like most of the known peptide-coding genes, si:zfos-2372e4.1 is highly expressed in a specific EEC cluster, namely cluster 5, which is also enriched for ccka expression (Fig 1E). Notably, in other zebrafish scRNA-seq datasets, expression of ccka is most highly correlated with expression of si:zfos-2372e4.1 (r = 0.74) [91], demonstrating a consistency of this pattern across laboratories. However, si:zfos-2372e4.1 had no gene or protein annotations. On our analysis, predictive software Signal P [92,93] assigned a high likelihood (0.81) that Si:zfos-2372e4.1 contains a signal peptide sequence at amino acid 1–23, indicating the protein is secreted. The prediction of a signal peptide is supported by the lack of detected peptides aligning to the N-terminus of the prepropeptide (Fig 2F).

To identify the potential functions of this gene, we performed a TBLASTN search using the full prepropeptide amino acid sequence, and the only similar sequences were in other fish species—largely in unannotated or predicted genes. Intriguingly, however, there was an approximately 20 amino acid sequence in Si:zfos-2372e4.1 that consistently aligned with those proteins in other species, suggesting that portion of the protein may be conserved and therefore perhaps functional. We looked specifically for any homology in gar, a holostean fish whose lineage diverged from teleost before the teleost-specific third round (3R) whole genome duplication (S6A Fig) [94]. Our search identified a gene in gar that, while also unannotated, allowed us to bridge to additional vertebrate species [95], including an elephant shark gene labeled as prosecretin with homology to SECRETIN (SCT) genes in humans and other tetrapods (S6B Fig) [96].

To further investigate the possibility that zebrafish si:zfos-2372e4.1 may be a SCT ortholog, we performed a syntenic analysis (S6B Fig). We identified genes in the ~250,000 base pair region upstream and downstream of the human SCT gene. Over half of those genes we identified had orthologs in zebrafish as annotated by Ensembl [97], and we traced the chromosomal arrangement of those genes in the Chondrichthyes elephant shark (Callorhinchhus milii), the Osteichthyes gar (Lepisosteus oculatus), and the Osteichthyes zebrafish (Danio rerio). While some rearrangements did occur, the immediate neighborhood remained remarkably consistent across all species (S6B Fig, S3 Table). This syntenic conservation across bony fish, cartilaginous fish, and tetrapods supports our identification of si:zfos-2372e4.1 as sct and suggests that this organization was present in the last common ancestor of all jawed vertebrates (S6A Fig) [98]. Given the evidence of homology from our TBLASTN and synteny assessments, we conclude that si:zfos-2372e4.1 is the zebrafish sct ortholog.

We aligned both known Secretin prepropeptides and unannotated proteins that were returned from our TBLASTN search (Fig 2G). We found the homology within fishes and between fishes and mammals in the region annotated as human secretin was relatively limited. However, we did detect several peptides aligning to this region in zebrafish (shaded in pink in Fig 2F). Moreover, we detected several additional peptides that aligned to a region that showed strong homology in fishes (shaded in gray in Fig 2F). All of these peptides had an N-terminal pyroglutamylated glutamine residue, supporting that these are genuine processed peptides, and not just degradation products. Functional studies are needed to test which, if any, of the zebrafish Secretin-derived peptides have secretin-like or potentially novel function. While we highlight Ghrelin- and Secretin-derived peptides as examples of conservation and potential novelty, we anticipate that the many additional zebrafish EEC peptides included in S4 Fig will serve as a valuable resource for future studies.

Zebrafish EEC subtypes show distinct expression patterns in vivo

To understand how the seven zebrafish EEC populations we identified develop and function in vivo, we assembled reporters for the EEC progenitor population and five of the EEC subtypes (Fig 3A): neurog3 and its transgene reporter neurog3:QF2 for cluster 0, ghrl and its transgene reporter ghrl:QF2 for cluster 1, peptide YYb (pyyb) and anti-PYY antibody for cluster 2, glucagon a (gcga) and its transgene reporter gcga:GFP for cluster 3, cholecystokinin a (ccka) and anti-CCK antibody for cluster 5, and transient receptor potential cation channel, subfamily A, member 1b (trpa1b) and its transgene reporter trpa1b:GFP for cluster 6. Transgenic reporters gcga:GFP and trpa1b:GFP and anti-PYY and anti-CCK antibodies were already available [99–102], but we developed new promoter-driven QF2 transgenic lines for neurog3 (using a 2114 bp neurog3 promoter) [103] and ghrl (using a 647 bp ghrl promoter) (S7 Fig and Methods). These reagents leverage the QF2-QUAS binary expression system, enabling us to combine our QF2 reporters with different QUAS tools [104,105]. As expected, our ghrl reporter co-labels with the neurod1 reporter and labels sparse EECs in larval and adult fish (S7E–S7I Fig). Our ghrl reporter also co-labels with cells that stain with anti-Ghrelin antibody and cells that label with ghrl mRNA using hybridization chain reaction (S7J–S7M Fig). Our neurog3 reporter likewise co-labels with the neurod1 reporter and labels a similar number of cells in the intestine as previously reported lines generated with the same promoter (S7S and S7T Fig) [106].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. In vivo imaging of EEC subtypes shows distinct spatial and temporal patterns during larval development.

(A) Summary of the hormones, transcription factors, and receptors significantly enriched in each EEC cluster and the reporter used to characterize that cluster. Bolding indicates genes from which we detected peptides. Parentheses indicate hormone-coding genes of interest expressed in that cluster that did not meet significance threshold for enrichment. Note, annotations of gast (see scRNA-seq methods), tac4b (see scRNA-seq methods, S15 Fig), and sct (see Results and Figs 2 and S6) were not previously identified and were defined in this paper. (B–S) Representative images from 6 days post fertilization (dpf) fish show subtype distribution and overlap with pan-EEC reporter neurod1. Individual cells are shown in inset. Quantitation of EEC subtype cell numbers from 3 to 6 dpf fish are also shown with each dot representing an individual fish. All scale bars are 100 μm. (T) 3 dpf, (U) 4 dpf, (V) 5 dpf, and (W) 6 dpf subtype localization data shown as frequency distributions of cells’ locations along the length of the gut divided into percentiles from proximal to distal. Distributions are smoothed for visualization. Underlying data can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003522.g003

To understand the dynamics of these EEC subtype reporters during intestinal development, we crossed each of them to the pan-EEC reporter neurod1:RFP [26] and imaged them daily from 3 to 6 days post fertilization (dpf) (S8 and S9 Figs). Cell counts from each day and representative images at 6 dpf are shown (Fig 3B–3S). For clarity, we refer below to cells labeled by a transgenic reporter or antibody marker as “+” for that respective gene (e.g., ghrl+). Subtype reporters showed a variety of prevalence and regional distribution patterns (Fig 3T–3W), which may contribute to their sub-specialized functions. ghrl+, PYY+, and CCK+ cells—labeling subtype clusters 1, 2, and 5, respectively—were all concentrated proximally within the intestine. Mammalian orthologs Ghrl/GHRL and Cck/CCK are also proximally skewed in the small intestine, but Pyy/PYY is more prevalent in the distal small intestine and colon in mammals [1,107]. gcga+ cells from subtype cluster 3 peaked in the midgut, consistent with the enrichment of the transcription factor osr2 in this cluster [72]. In accordance with mammalian literature, trpa1b+ cells labeling subtype cluster 6 (enterochromaffin cells) were the most prevalent subtype and were relatively evenly distributed along the length of the intestine [108]. To further explore the regional distribution of EEC subtypes, we used Seurat module scores to look at the expression of regionally enriched genes [109]. Consistent with our imaging results, midgut genes were enriched in cluster 3 and anterior gut markers were enriched across many subtype clusters (S10 Fig). Overall, the subtype reporters appeared as early as 3 dpf but became more widely expressed at 4 dpf, in alignment with neurod1 reporter activity, growth of the intestine, and mRNA in situ hybridization data [29,30]. After 4 dpf, EEC subtype numbers increased with age except for gcga+ and ghrl+ cells, which showed a trend towards decreasing after their initial increase in cell number at 4 dpf.

Importantly, while we apply these antibodies and reporters as useful markers of distinct EEC populations, EEC subtype definitions are operational and can vary across modalities and resolutions. For example, while expression of sst2 and ghrl is highest in cluster 1 (S1 Table), cluster 1 contains cells that express only sst2, cells that express only ghrl, cells that express both ghrl and sst2, and cells that express neither ghrl nor sst2 (S7N–S7R Fig). With greater clustering resolution and sequencing depth, these populations could be considered distinct subtypes. In this study, we operationally define cluster 1 as an EEC subtype and use the ghrl reporter as a label for this subtype.

Ablation of neurog3+ cells leads to reduction in EECs limited to select subtypes

We predicted that cells in scRNA-seq cluster 0 were EEC progenitors based on their enriched expression of neurog3 and other EEC progenitor transcription factors and their early position on the pseudotime-predicted differentiation trajectory (Fig 1G and 1H). As cluster 0 made up a relatively small proportion of total EECs by scRNA-seq (Fig 1B), we were surprised to see our neurog3 reporter label so many EECs in vivo (Fig 3B–3D). It is unclear if this is due to promiscuity of the neurog3:QF2 transgene or perdurance of protein expressed from the transgene reporter. The mouse ortholog, Neurog3, is known to have transient expression in EEC progenitors [110], thus the perdurance of both QF2 and GFP may contribute to a long-lasting labeling of cells that transiently expressed neurog3 in their developmental past. Thus, the unexpectedly high number of neurog3+ cells at 6 dpf may indicate that a larger portion of EECs transiently express neurog3 at earlier developmental stages.

To test this hypothesis, we used a combination of reporters to genetically induce Diphtheria toxin subunit A (DTA) expression, and thus cell death, in any neurog3+ cell in the intestine (Fig 4A). Specifically, we leveraged a transgene reporter that uses gata5 regulatory sequences to drive intestinal epithelial expression of a loxP-flanked mCherry upstream of DTA (gata5:RSD) [27]. When we combined this reporter with our neurog3:QF2 and QUAS:Cre alleles, the mCherry is recombined out in neurog3+ cells, leading to expression of DTA and cell death, which we operationally call “neurog3+ cell ablation.” This strategy affects cluster 0, which has enriched neurog3 expression, and effectively prevents the emergence of any further mature subtypes that ever pass through a neurog3+ state.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Selective ablation of neurog3+ cells in the intestine reduced total EECs and several EEC subtypes.

(A) Schematic of neurog3+ cell ablation. (B) Schematic of regional scoring system. Total and regional counts of (C) neurod1+ cells, (D) ghrl+ cells, (E) PYY+ cells, (F) gcga+ cells, (G) CCK+ cells, and (H) trpa1b+ cells in control and neurog3+ ablated fish are shown. Each dot represents a 6-day post-fertilization fish. Statistical significance was calculated by unpaired t test for total cell numbers and by two-way ANOVA for regional analysis. Significance annotations are as follows: ns (p > 0.05), * (p < 0.05), ** (p < 0.01), *** (p < 0.001), **** (p < 0.0001). Underlying data can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003522.g004

We evaluated the total numbers and regional distribution of neurod1+ cells (Fig 4B) in this neurog3+ cell ablation model and found that roughly 50% of all EECs are lost with the effect being more pronounced in the proximal gut (Fig 4C). A previous study showed that expression of the gata5:RSD transgene is relatively higher in the proximal region of the larval intestine and that combination with neurod1:Cre (neurod1+ cell ablation) led to ablation of EECs in the proximal and mid intestine while retaining some EECs in the most distal region [27]. To test if this regional activity of the gata5:RSD transgene contributed to our observed patterns of regional EEC loss following neurog3+ cell ablation, we performed a series of control experiments. First, we quantified the number of neurog3+ cells in the neurog3-ablated fish and found that while almost no neurog3+ cells remained in the first three quarters of the intestine (qrt1–3), levels of neurog3+ cells were comparable to unablated control fish in the fourth quarter of the gut (qrt4; S11A and S11B Fig). We then likewise evaluated the effects of neurod1 ablation on neurod1+ cells and observed a similar near complete loss of neurod1+ cells in qrt1–3 and preservation of neurod1+ cells in qrt4, consistent with previous studies [27]. In comparing the neurod1+ cell numbers in control, neurog3-ablated, and neurod1-ablated fish, there were significantly fewer neurod1+ cells in the proximal three-quarters of the gut in the neurod1-ablated fish as compared to the neurog3-ablated fish (S11C Fig). Similarly, trpa1b+ cells, while preserved in qrt4 in all animals, were significantly more reduced in neurod1-ablated than neurog3-ablated fish (S11D Fig). Together, these experiments show that combining a Cre driver with gata5:RSD efficiently kills cells in the first three quarters of the larval gut. They furthermore demonstrate that neurog3+ ablation significantly reduces EEC numbers, but not to the same extent as neurod1 ablation, supporting the conclusion that, in contrast to mammalian systems [10], a neurog3+ cell state is not required for all EEC development in zebrafish. These results also demonstrate that ablation using the gata5:RSD transgene is inefficient in the final quarter of the gut, likely due to lower gata5 expression in that region. To account for that lack of ablation in qrt4, we calculated total EEC numbers in our ablation experiments both including qrt4 (Figs 4C–4H and 5B–5F) and excluding qrt4 (S11E and S11F Fig) and observed the effect of ablation was the same regardless of whether qrt4 cells were included or not.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Selective ablation of ghrl+ cells suggests lineage relationships between ghrl+ cells and other EEC subtypes.

(A) Schematic of ghrl+ cell ablation. Total and regional counts for (B) neurod1+ cells, (C) PYY+ cells, (D) gcga+ cells, (E) CCK+ cells, and (F) trpa1b+ cells in control and ghrl-ablated fish are shown. Each dot represents a 6-day post-fertilization fish. (G) Schematic of ghrl+ cell lineage tracing. (H) Representative images of ghrl lineage tracing with PYY staining showing rare double-positive cells. Scale bars are 100 μm. Statistical significance was calculated by unpaired t test for total cell numbers and by two-way ANOVA for regional analysis. Significance annotations are as follows: ns (p > 0.05), * (p < 0.05), ** (p < 0.01), *** (p < 0.001), **** (p < 0.0001). Underlying data can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003522.g005

We next combined our cell ablation strategy with EEC subtype reporters to ask if certain EEC subtypes were particularly affected by neurog3+ cell ablation. Because we could not combine the neurog3:QF2 and ghrl:QF2 reporters, we generated a neurog3:Cre transgenic line to test the effects of neurog3+ cell ablation on ghrl+ cells. We confirmed that neurog3+ cell ablation induced with the neurog3:Cre reporter similarly reduced neurod1+ cell numbers as compared to the neurog3:QF2; QUAS:Cre reporter (S12 Fig). When evaluating EEC subtype populations in the context of neurog3+ cell ablation, we found that PYY+, CCK+, and trpa1b+ cells were all reduced by roughly 30% while gcga+ and ghrl+ cells were unaffected (Fig 4D–4H). Consistent with our reporter characterization (Fig 3), reductions in EEC subtypes were observed in the region of the gut where they were most prevalent—proximally for PYY+ and CCK+ cells and distally for trpa1b+ cells. This suggests that a portion of PYY+, CCK+, and trpa1b+ subtypes pass through a neurog3+ cell state during their differentiation.

To test if the effects of neurog3+ ablation were related to the loss of the neurog3 gene product, we quantified the number of EECs in neurog3 mutants [29] and control animals and observed no difference (S13D Fig). This suggests that the neurog3 gene is not required for zebrafish EEC differentiation, which is consistent with what has been reported in zebrafish pancreatic islet differentiation [29], but contrasts with the known essential role of Neurog3 in mammalian EEC development [10]. In combination with the results of the neurog3+ ablation experiments, we conclude that many zebrafish EECs pass through a neurog3+ cell state, but do not require neurog3 function for their overt development.

Ablation of ghrl+ cells leads to reduction in EECs in subtypes distinct from neurog3+ ablation

The persistence of ghrl+ cells despite neurog3+ cell ablation is reminiscent of mammalian data that shows Ghrl expression and Ghrl+ cell numbers were maintained or increased upon deletion of several transcription factors important for EEC differentiation, such as Nkx2.2a and Pax4 [19,65,66,111–114]. In addition, our pseudotime analysis predicted a differentiation trajectory moving through ghrl+ cells, strikingly similar to several mammalian scRNA-seq datasets that suggested a Ghrl+ EEC progenitor population [23–25,115]. Older lineage tracing data also supports a role for Ghrl+ cells as progenitors of other endocrine populations in the stomach and pancreatic islet [22]. However, the hypothesis that Ghrl+ cells in the intestine give rise to other EEC subtype populations remained untested.

Using a similar genetic approach as above, we used gata5:RSD to selectively kill any ghrl+ cell in the intestinal epithelium, which we operationally define as “ghrl+ cell ablation” (Fig 5A). Despite ghrl+ cell numbers peaking at approximately 15 cells per animal during larval development (Fig 3G), we found ghrl+ cell ablation resulted in a loss of roughly 40 EECs, indicating other EEC subtypes might be affected (Fig 5B). Indeed, we found a 30% reduction in PYY+ and gcga+ EECs in ghrl-ablated fish while CCK+ and trpa1b+ cell numbers remained unaffected (Fig 5C–5F). Notably, this is distinct from the neurog3+ ablation result where PYY+, CCK+, and trpa1b+ cells were reduced and ghrl+ and gcga+ cell numbers were unaffected (Fig 4). These data suggest a neurog3-independent role for ghrl+ cells in the differentiation of PYY+ and gcga+ EEC subtypes. This could be explained by PYY+ and gcga+ cells passing through a ghrl+ cell state or by the loss of a secreted factor from ghrl+ cells that promotes differentiation of these subtypes.

To distinguish these possibilities, we first used lineage tracing tools to indelibly label a cell with GFP if it has ever expressed ghrl (Fig 5G). We then performed PYY staining and found rare occurrences of double-positive cells, confirming that some PYY+ cells have indeed passed through a ghrl+ cell state (Fig 5H). This suggests that the impact of ghrl+ cell ablation on PYY+ cells is at least in part cell autonomous. Due to overlapping fluorophores, we were not able to perform a similar study with gcga+ cells.

Given the small number of ghrl-derived PYY+ cells in our lineage tracing analysis, we hypothesized that ghrl+ cells could also be regulating other EEC subtypes non-cell autonomously through secreted products. As ghrelin peptides were prevalent in our peptidomics dataset (Fig 2D) and have been shown to stimulate growth of intestinal cells [116,117], we tested if loss of the ghrl gene impacted EEC development. We used CRISPR/Cas9 to generate ghrl mutant zebrafish and evaluated the effects on neurod1+ EECs. We found that EEC numbers were equivalent across ghrl mutant and wildtype fish (S13 Fig), suggesting that loss of ghrl gene function does not overtly impact EEC development.

Ablation of neurod1+ cells, but not ghrl+ or neurog3+ cells, impacts adult survival and growth

To test the ramifications of each of our EEC ablation models on zebrafish growth and survival, we raised ablated fish with their sibling controls to adulthood. We maintained at least five tanks of each control and ablated condition for each experimental replicate and checked survival at several timepoints. At the end point of each study, we also randomly selected fish in each group for measurement of standard length (S14A and S14B Fig) as a metric of animal growth [118]. As we anticipated, neurod1+ ablated fish showed the most severe phenotype, with evidence of reduced survival as early as 2 weeks post fertilization (S14C Fig). Additionally, at the 6-week endpoint, neurod1+ ablated fish were also observed to be smaller than their sibling controls (S14C Fig). In contrast, neither neurog3+ nor ghrl+ ablated fish showed survival or growth defects over 6 weeks (S14D, S14F Fig). To further interrogate any possible differences, we raised a second set of neurog3+ and ghrl+ ablated and control fish to 12 weeks post fertilization, a commonly accepted time point for full maturity [118]. However, we still did not observe any differences in survival or growth (S14E, S14G Fig), suggesting neurog3+ and ghrl+ ablation have no significant impact on feeding behavior and adult development.

Ethics statement

All zebrafish studies were approved by the Institutional Animal Care and Use Committees of Duke University (protocol A061-22-03) and Michigan State University (protocol PROTO202200367).

Zebrafish lines and husbandry

Zebrafish stocks were maintained on an Ekkwill (EK) background on a 14-hour/10-hour light/dark cycle at 28.5°C in a recirculating aquaculture system (Pentair). From 6 to 14 dpf, larvae were fed Zeigler AP100 <50-micrometer larval diet (Pentair, LD50-AQ) twice daily. Beginning at 14 dpf, larvae were fed Artemia (Brine Shrimp Direct, BSEACASE) once daily, supplemented with a daily feed of Skretting Gemma Micro 75 (Bio-Oregon, B5676). From 28 dpf, the Gemma Micro 75 diet was replaced with Gemma Micro 300 (Bio-Oregon, B2809). At the onset of breeding age or sexual maturity, adult fish were given a 50/50 mix of Skretting Gemma Micro 500 (Bio-Oregon, B1473) and Skretting Gemma Wean 0.5 (Bio-Oregon, B2818) and one feeding of Artemia daily.

For experiments involving zebrafish larvae, adults were bred naturally in system water and fertilized eggs were transferred to 100 mm petri dishes containing ~60 mL of egg water at approximately 6 hours post-fertilization. The resulting larvae were raised under a 14-hour light/10-hour dark cycle in an air incubator at 28°C at a density of 1 larvae/mL water. All the experiments performed in this study ended at 6 dpf unless specifically indicated. As sex determination in zebrafish occurs after 6 dpf, we do not report numbers of male or female fish in each experiment. The following engineered zebrafish lines were used in this study:

TgBAC(cldn15la:EGFP)^pd1034 [167], Tg(neurod1:TagRFP)^w69 [168], TgBAC(trpa1b:EGFP)^a129 [102], TgBAC(gata5:loxp-mcherry-stop-loxp-DTA)^pd315 [27], TgBAC(neurod1:EGFP)^nl1 [169], Tg(gcga:EGFP)^ia1 [170], Tg(β-actin2:loxP-mCherry-loxP-GFP)^pd31 [171], Tg(QUAS:GFP; cryaa:mCherry)^pd1199 [172], neurog3^sa211 [29], Tg(neurog3:QF2; cmlc2:GFP)^rdu107 (generated in this study), Tg(ghrl:QF2; cmlc2:GFP)^rdu108 (generated in this study), Tg(QUAS:cre; cryaa:mCherry)^rdu109 (generated in this study), Tg(neurog3:Cre; cmlc2:GFP)^rdu110 (generated in this study), ghrl ^rdu111 (generated in this study), and Tg(QUAS:TagBFP)^rdu113 (generated in this study).

Construction of zebrafish lines

The Gateway Tol2 cloning approach was used to generate Tg(neurog3:QF2; cmlc2:GFP), Tg(ghrl:QF2; cmlc2:GFP), Tg(QUAS:cre; cryaa:mCherry), and Tg(neurog3:Cre; cmlc2:GFP) plasmids [103]. The 2114 base pair neurog3 promoter was previously reported [106] and generously shared by Dr. Olov Andersson (Karolinska Institutet) and cloned into a p5E-MCS vector [103] using TAKARA In-Fusion Snap Assembly master mix (Takara, 638947) with HindIII-linearization of the backbone and amplification of the insert with primers CGGTATCGATAAGCTCCGCGGCCGCCGTACTCGA and ATTCGATATCAAGCTCGGCGCGCCCACCCTTTCTGTA. The pME-QF2 [105], p5E-QUAS [173], p3E-polyA [103], pDestTol2CG2 [103], and pDESTtol2pACrymCherry [174] plasmids were previously reported and obtained from Addgene. The pME-Cre plasmid was previously reported [175] and generously shared by Dr. Mark Cronan. The 647 base pair ghrl promoter was amplified from purified (Wizard Genomic DNA purification kit, A1120) EK DNA using primers cttaagcttTTCAGAATTCATATCAGATCACAGACACT, cttccgcggCACCTTAGTCTTAATTCTTTGCTACATAC. It was then cloned into the p5E-MCS [103] via digestion of insert and backbone with HindIII (NEB, R3104S) and SacII (NEB, R01575S) followed by phosphatase treatment (NEB, M0289S) and ligation (NEB, M0202S). The appropriate p5E, pME, and p3E entry vectors were cloned into destination vectors with either a green heart (cmlc2:GFP; pDestTol2CG2) or red eye (cryaa:mCherry; pDESTtol2pACrymCherry) as listed above using an LR-Clonase (ThermoFisher, 12538-120) reaction.

The ghrl mutant line was generated using CRISPR-Cas9. The guide RNAs (gRNAs) were designed using the “CRISPRscan” tool (www.crisprscan.org/) [176] and purchased fully synthesized from Integrated DNA Technologies (see S4 Table). At the one-cell stage, EK strain zebrafish embryos were injected with 1–2 nL of a cocktail consisting of Cas9 protein (1,000 ng/μL), gRNA (150 ng/μL), and 0.05% phenol red. Injected embryos were screened for mutagenesis with forward primer CACAGTGACAGTTGTAGACTTTAATGCTAAT and reverse primers GTCTCTAAGAAGATTCTCCAGAAGATTCTGA (mutant), AAACATGCTGCTGGCACGGCA (wildtype). The mutations were further determined through Sanger sequencing of the region encompassing the gRNA targeting sites. Ten F0 injected fish were out-crossed to wild-type fish and progeny were screened to test if mutations were transmissible. F1s from F0 founders were then raised to adulthood and sequenced. Two F1s were found to have identical lesions—the deletion started after the first nucleotide in the first exon and continued through the first 17 nucleotides of the third exon—and were bred to generate a stable line. F2s and F3s from this line were used in the reported experiments and were given the allele name rdu111. When testing ghrl mutants, genotyping of the mttp locus was used as a control to verify there was viable DNA in samples using forward primer AGAGACGGTGTCCAAGCAGG and GCTCAAAGACTTTCTTGC with an expected band of 137 base pairs [177].

mVISTA alignment

To identify conserved regions of the ghrl promoter across Danio species, we exported the ghrl sequence and annotation from ensembl.org [97]. We blasted the D. rerio Ghrl amino acid sequence against other Danio species and extracted 5 kb upstream and downstream of the aligned region for each species. We then put each of these sequences into the mVISTA alignment [178,179], using the D. rerio sequence and annotation as the reference. We identified and amplified a 647 base pair region upstream of the ghrl transcriptional start site that was highly conserved across Danio species (S7B Fig).

Larval immunofluorescence and live imaging

All larval imaging was performed on Andor Dragonfly Spinning Disc Confocal plus with a 20× lens in the Duke Light Microscopy Core Facility. For live imaging experiments, zebrafish larvae were anesthetized with tricaine and mounted in 1% low-melt agarose. For larvae imaged at 3 and 4 dpf, the yolk was mechanically removed with fine watchmaker forceps prior to mounting.

Whole-mount immunofluorescence staining was performed as previously described [26]. Larvae were euthanized with 5× tricaine and roughly 20 larvae per condition were transferred to a 1.5 mL tube. Excess media was removed and 1 mL of chilled 4% paraformaldehyde was added. Larvae were incubated in fixative overnight (>16 hours) at 4°C. Samples were then washed 2 × 5 min in PBS. For samples of 3 or 4 dpf larvae, yolk was then mechanically removed using fine watchmaker forceps before dehydrating in 50% methanol/50% PBS solution for 5 min. Larvae were then washed 3× in 100% methanol before incubation in 100% methanol at −20°C for at least 2 hours. Larvae were then rehydrated in serial 50% methanol/50% PBS, 100% PBS washes, and permeabilized at room temperature for at least 20 min with PBS with 0.1% Tween-20, 1% DMSO, and 0.3% Triton X-100. Larvae were then washed and blocked with 4% bovine serum albumin solution for at least 30 min at room temperature before washing and incubating in primary antibody solution for at least 24 hours at 4°C. Primary antibodies in this study were diluted at 1:100 [rabbit anti-PYY (custom antibody generated in [100]), rabbit anti-CCK (custom antibody generated in [99]), rabbit anti-Ghrl (Anaspec, 55529), goat anti-IgG (Jackson ImmunoResearch, 115-005-044)] or 1:500 [chicken anti-GFP (Aves, GFP-1010)]. Larvae were then washed every 15 min for 2 hours before incubation with Alexa Fluor Invitrogen secondary antibodies diluted at 1:250 [488 goat anti-chicken (ThermoFisher, A11039), 647 goat anti-rabbit (ThermoFisher, A21244), 568 goat anti-rabbit (ThermoFisher, A11011)]. Larvae were incubated with secondary antibodies overnight (>16 hours) at 4°C, washed every 15 min for 1 hour, and mounted for imaging.

neurod1 cell analysis in neurog3 mutants

The neurog3^sa211 mutant allele was obtained from the Zebrafish Mutation Project (Sanger Institute, UK). Adult heterozygous neurog3^sa211 zebrafish were bred naturally in system water, and fertilized eggs were collected, transferred to 100 mm petri dishes containing embryo medium, and allowed to develop at 28.5°C. All steps of the staining were performed on nutator at room temperature except when indicated otherwise. Approximately 30 larvae were fixed in 750 µL of 4% PFA prepared in 1× sweet buffer (8% sucrose, 0.2 M CaCl₂, 0.2 M PO₄ buffer, pH 7.3) for 2 hours. Then, larvae were washed twice in 0.5% Phosphate buffer saline-Triton X-100 (PBSTX) for 15 min, and stored in PBSTX at 4°C. Prior to blocking, larvae were washed 5× in double distilled water for a minimum of 60 min each and then they were incubated for 1 hour in blocking solution [0.5% PBS-Triton X-100, 1% DMSO, 2% bovine serum albumin (Sigma-Aldrich, A3059), 5% normal growth serum (ThermoFisher, PCN5000). Next, larvae were incubated for 16–18 hours in anti-DsRed primary antibody (1:300; Takara Bio, 632496) diluted in blocking solution. Larvae were then washed 3× for a minimum of 3 hours each in 0.3% PBS Triton X-100 before incubating for 16–18 hours in Alexa Fluor 568 secondary antibody (1:1,500; ThermoFisher, A11035) diluted in blocking solution. Finally, larvae were washed 3× for a minimum of 3 hours each in dark and stored at 4°C in 0.3% PBS Triton X-100 before mounting.

Before imaging the immunostained intestines, each larva was tail clipped for genotyping and the body stored in 0.3% PBSTX at 4°C. For extraction of genomic DNA, the fin tissue was incubated in 25 µL of 1× lysis buffer [per 1 mL: 100 µL of 10× PCR buffer (BioLabs NEB, B9004S), 50 µL ∼20 mg/mL of proteinase K (ThermoFisher, 100005393), and 850 µL of nuclease-free water] at 55°C for 50 min, followed by 98°C for 10 min. PCR amplification was performed using 1 µL of genomic DNA, 19 µL master mix containing 10 µL of 2× GoTaq Green Master Mix, 7 µL of nuclease-free water, and 1 µL each of the forward (5′-CAAGATCCTCGTGCGCTTTG-3′) and reverse (5′-CCTGATCCTGGGCGTGATTT-3′) primers. The PCR protocol was as follows: 95°C for 2 min; 34 cycles of 95°C for 30 s, 66°C for 30 s, and 72°C for 60 s; final extension at 72°C for 5 min; hold at 12°C. PCR products were cleaned using ExoSAP-IT following the manufactures protocol (ThermoFisher, 78201.1.ML), then sent for Sanger sequencing. Larvae were classified as homozygous wild type or homozygous mutant based on the absence or presence of the mutant allele (ZMP, Zebrafish Mutation Project, Zebrafish Mutant Resource) [180,181]. Only intestines of homozygous wild-type or homozygous mutant were mounted and imaged.

For quantification of EECs, intestines were dissected with forceps and mounted on glass slides in 0.3% PBSTX or Fluoromount-G (SouthernBiotech, 0100-01) and covered with a coverslip. Confocal z-stacks images were captured using the 3i spinning disk confocal microscope with a 20× objective, operated through Slidebook6 software (Version 6.0.13).

In situ hybridization chain reaction

Probes for ghrl RNA (S4 Table) were designed according to previously published parameters [182]. cubn probes were generously shared by Dr. Michel Bagnat. Hairpins and amplification, hybridization, and wash buffers were purchased from Molecular Instruments [183]. Our methods were adapted from previously published procedures in zebrafish [109,182]. 6 dpf larvae were fixed in 4% PFA for 3 hours at room temperature, followed by PBS and ice-cold acetone washes before being incubated in acetone at −20°C for 8 min. Larvae were then washed with PBS and incubated with Molecular Instruments probe hybridization buffer for 30 min rotating at 37°C. Larvae were then incubated in probe solution (8 nM of each probe) for 48 hours rotating at 37°C. To remove excess probes, larvae were washed 4 × 30 min each in wash buffer pre-warmed to 37°C followed by 2 × 10 min washes in 5× SSCT buffer at room temperature. Larvae were then incubated in amplification buffer for 30 min at room temperature while hairpins were prepared. Sixty pmol of each hairpin were heated separately to 95°C for 90 s and then snap cooled in the dark at room temperature for 30 min. Larvae were incubated in hairpin solution rotating in the dark overnight at room temperature. Excess hairpins were removed with 2 × 5 min, 2 × 30 min, and 1 × 5 min washes in SSCT buffer at room temperature. Larvae were then mounted and imaged with spinning disc confocal.

Image analysis with ImageJ

Images were processed and analyzed in ImageJ [184]. Images were automatedly blinded such that the scorer did not know the condition (e.g., ablated or control) of each image. The start and end of the gut was manually marked using the Cell Counter plugin and x, y coordinates were extracted. neurod1+ cells were automatedly counted from maximum intensity projections using the Analyze Particles function with minimum object size of 10 pixels². The number and location of cells was automatedly extracted in R and locations were zero-ed using corresponding manual gut markers. To account for overlapping cells being grouped into a single object, the median object size was used to assign a weight to each object. Objects twice as large as the median size given a weight of 2, objects three times as large as the median were given a weight of 3, and so on, with a minimum weight of 1. Weights were summed to count the total number of cells in an image.

neurog3+ cells were counted with a similar automated approach. A composite of neurod1 and neurog3 channels was used to identify objects using the Analyze Particles function and the fluorescence of each channel was extracted for each identified object. Objects with a ratio of neurog3:neurod1 signal >2 were counted as neurog3+ only, objects with a ratio of neurog3:neurod1 signal <0.5 were counted as neurod1+ only, and all objects with a ratio in between 2 and 0.5 were counted as double positive. The same adjustment for overlapping cells described above was also applied.

Because of increased background of various sources causing complications with automated thresholding and counting, images from gcga, trpa1b, and ghrl reporters along with PYY and CCK stains were manually counted using the Cell Counter plugin. Counts for reporter characterization in Fig 3 were taken for each slice of the captured image while maximum intensity projections were used for ablation image analysis for higher throughput. In all cases, the number and location of each manually counted cell were extracted for further analysis.

In addition to total counts, the location of automatedly or manually counted cells was determined relative to the manually annotated start and end points of the gut. In cases where a single object was given a weight >1 and counted as multiple cells, the x, y coordinates of the original object was used for each. Cells were binned into four evenly divided bins along the length of the intestine for visualization of counts per quarter of the gut in bar plots (see Figs 4, 5, S9, and S11) or into 100 evenly divided bins for visualization of density per percentile of the gut in density plots (see Fig 3T–3W).

Adult sections

Adult zebrafish 3–4 months post fertilization of either sex expressing Tg(neurod1:RFP); Tg(ghrl:QF2; cmlc2:GFP); Tg(QUAS:GFP; cryaa:mCherry) were retrieved from the circulating aquaculture system before morning feeding and transferred to a separate tank to fast before dissection. Several hours later fish were euthanized in 0.090% 2-Phenoxyethanol and dissected to retrieve the intestine. The intestine was fixed in 4% PFA overnight at 4°C then washed 3× the next day with cold 1× PBS. The intestine was mounted in 4% low-melt agarose in cryomolds and cured at 4°C for 1 hour. 200 μm cross sections of the tissue were sectioned via Vibratome (Leica VT1000S) and mounted on slides with Vectashield containing DAPI (Vectashield H-1200). Z-stack images were collected on a Zeiss 780 inverted confocal microscope at either 20× or 40× magnification. Image processing was done in FIJI including maximum intensity projections of z-stacks from each channel.

Adult Swiss rolls

Adult zebrafish (approximately one year post fertilization) expressing Tg(neurod1:RFP); Tg(ghrl:QF2; cmlc2:GFP); Tg(QUAS:GFP; cryaa:mCherry) were retrieved from the recirculating aquaculture system and euthanized with 0.090% 2-Phenoxyethanol. Intestines were dissected and flayed open to discard luminal debris using ice-cold 1× PBS. Once cleaned, a few drops of Carnoy’s fixative was used to stiffen the tissue prior to rolling and immediately washed with ice-cold 1× PBS. Intestine was rolled from the proximal to distal end, pinned through with a dissection pin (Fine Science Tools, 26002-10) to maintain shape, then transferred to 4% PFA overnight at 4°C. After overnight fixation, the tissue was transferred to a 30% sucrose solution with 0.02% sodium azide and stored at 4°C until tissue sank in preparation for cryosectioning. Intestines were then mounted in cryomolds (TissueTek, 4566) with FS 22 Clear (Leica, 3801480), frozen in a dry ice bath, and stored as blocks at −80°C until sectioning. Blocks were allowed to equilibrate to internal cryostat temperature for 1 hour prior to sectioning and 10 µm sections were collected via cryostat (Leica, CM1860). Slides were stored at −80°C until imaging. For imaging, sections were mounted with Vectashield containing DAPI (Vectashield, H-1200) and glass coverslips and imaged on an Andor Dragonfly confocal at 20x with z-stack and stitching.

Survival assays

neurod1, neurog3, or ghrl-ablated larvae were generated alongside sibling controls, which were identical to their ablated counterparts but lacked the Cre transgene. Groups of 25–30 larvae were placed in housing tanks and put on a recirculating aquaculture system at 6 dpf and were fed and maintained according to standard husbandry methods described above. The number of fish remaining in each tank was counted at three evenly spaced timepoints (2, 4, and 6 weeks post-fertilization for initial experiments and 4, 8, and 12 weeks for extended experiments for neurog3 and ghrl ablated conditions). Any tank that had >50% death rate at the first measurement was excluded from analysis as occasional tank die off is known to occur sporadically. Specifically, excluded tanks were as follows: one control tank in the 12-week neurog3 ablation experiment and one control and two ablated tanks in the 12-week ghrl ablation experiment. At the last time point of each experiment (6 or 12 weeks), manual standard length measurements of roughly representative fish from each tank were performed.

Single-cell RNA sequencing analysis

We imported data from previously published adult [49] and larval [48] zebrafish intestinal scRNA-seq datasets into R (version 4.3.1). Using established markers of intestinal secretory cells, we selected just secretory cells from each dataset (S1 Fig) and integrated them using the SCT method in Seurat version 5.0.3 [185]. We further subsetted this dataset for just secretory progenitors and EECs and re-integrated and clustered it to generate the dataset shown in Fig 1B and analyzed in this manuscript. This final dataset included 2069 cells from the adult dataset and 1891 cells from the larval dataset. Multiple resolutions were evaluated and a resolution of 0.4 was ultimately used (S16 Fig). We used monocle3 (version 1.3.4) to perform pseudotime analysis where use_partition was set to TRUE and root_cells were set to cluster 7 [58]. Module scores were calculated using the AddModuleScore() function. The list of genes included in the regional modules score was derived from the enriched genes of the pharynx, anterior enterocyte, anterior LRE, posterior LRE, and cloaca clusters in [109] and are listed in the Supplemental Materials (S10 Fig, S1 Table). The integrated secretory progenitor and EEC dataset is available on Zenodo at at DOI 10.5281/zenodo.17342282.

Of note, the zebrafish gastrin gene is currently named either LOC100536965 or CR556712.1 in danRer11 and is annotated in Ensembl as a noncoding lncRNA (ENSDARG00000117418) for unknown reasons. However, multiple phylogeny studies have identified the exact genetic coordinates of LOC100536965/CR556712.1 as zebrafish gastrin [186–188] and peptides we detected derived from this gene align with RefSeq XM_021479754.1, which is a coding prediction. We therefore refer to the gene as gastrin. When searching for expression of gastrin in our scRNA-seq dataset, we found expression under the gene name CR556712.1 in the larval cells, but no expression of either gene name in the adult cells. Searching using either gene name in the complete published adult dataset [49] also did not return any results, suggesting the gene is either not annotated in that dataset, annotated with a different name, or else not expressed. Detection of gastrin-derived peptides in our adult peptidomics samples (S2 Table) suggests that gastrin is indeed expressed in adult EECs and that incomplete gene annotation may contribute to this discrepancy.

Further, what we annotated as tac4b is currently named si:ch211-131k2.2 in danRer11 (Ensembl ID ENSDARG00000096645). One of the listed protein names for its corresponding Uniprot entry (R4GDR9), however, is Protachykinin-1 isoform X. To evaluate if this gene may be a member of the tachykinin family, we used Ensembl and NCBI to identify orthologs of zebrafish tachykinin genes (tac1, tac3a, tac3b, tac4, si:ch211-131k2.2) in common carp (Cyprinus carpio carpio), catfish (Ictalurus punctatus), rainbow trout (Oncorhynchus mykiss), spotted gar (Lepisosteus oculatus), elephant shark (Callorhinchus milii), pig (Sus scrofa), mouse (Mus musculus), and human (Homo sapiens). We then performed a protein alignment and phylogenetic analysis using MEGA 12 to produce the phylogenetic tree shown in S15 Fig. This tree showed two tac4 paralogs in all the teleost fish species included in our analysis, but only one tac4 gene in gar. This pattern is commonly seen due to the third round (3R) of whole genome duplication in the teleost lineage (S6A Fig) [189,190]. Altogether, our analysis suggests that si:ch211-131k2.2 is a TAC4 ortholog. While prior studies had suggested zebrafish has two tac4 paralogs [191], gene annotation has remained obscure. Based on this alignment, we provisionally renamed zebrafish tac4 as tac4a and si:ch211-131k2.2 as tac4b.

To test the generalizability of our EEC subtype annotations, we downloaded scRNA-seq data from [56] available on the Broad Institute Single Cell Portal. Using the same markers described above, we identified an EEC cluster and subsetted and visualized it using UMAP dimensionality reduction to identify EEC subtypes (S3 Fig).

Statistical analysis

For the scRNA-seq analysis, statistical analyses for determination of the cluster-enriched markers were calculated using the FindConservedMarkers() function of the Seurat package in R with a Wilcoxon rank sum test. For all other experiments, statistical analysis was performed using unpaired t test, or one-way or two-way analysis of variance (ANOVA) with Tukey’s multiple comparisons test with GraphPad Prism. P < 0.05 was defined as statistically significant.

EEC FACs sorting

All media were made fresh the morning of the experiment except for the 6M guanidine hydrochloride, for which the same solution was used throughout the two-week collection period. For larval samples, Tg(neurod1:RFP); TgBAC(cldn15la:GFP) adult breeders were crossed and embryos were collected at roughly 6 hours post-fertilization and maintained in egg water [192] at approximately 1 embryo/mL density. Unfertilized embryos were removed at 1 dpf and larvae were anesthetized and sorted for presence of both reporters at 5 dpf. The GentleMACS disassociator (Miltenyi, 130-093-235) was placed in the 4°C room to come to temperature for the following day’s experiment. At 6 dpf, 75 larvae were euthanized in tricaine and pooled into a single sample with 9 samples being processed concurrently with single positive and non-transgenic controls. Larvae were transferred with minimal media to Gentle MACS C tubes (Miltenyi 130-096-334) containing 2 mL of freshly made disassociation buffer [10 mg/mL cold protease (Sigma, P5380-1G), 10 μM ROCK inhibitor (VWR, S1049), 2.5 mg/mL DNase (Sigma, DN25) in Dulbecco’s PBS]. Samples were processed 5 times with protocol C_01 on GentleMACS disassociator with 10 min of incubation on shaker at 4°C after each round. Ten mL of FACS media [0.1% BSA (Fisher, BP1600-100), 10 μM ROCK inhibitor (VWR, S1049) in HBSS (Sigma, H9394-1L)] was added and the contents were poured through a 30 micrometer strainer (Miltenyi 130-098-458) into a fresh 50 mL conical tube. The strainer was washed with an additional 10 mL of FACS media and samples were spun down at 250g for 5 min at 4°C, decanted, and resuspended in 1 mL of FACS media. Samples and controls were then transferred to FACS tubes (Corning, 352052) with 5 μL of 7AAD (Sigma, A9400-1MG) to stain dead cells. Cells were then immediately subjected to FACS at the Duke Cancer Institute Flow Cytometry Shared Resource. neurod1:RFP; cldn15la:GFP double positive, 7AAD negative cells from each sample were sorted into a single 1.5 mL LoBind tube (Sigma, Z666505) with 500 μL of 6M guanidine hydrochloride for a total of roughly 30,000 cells. This constituted a single sample for downstream peptidomic analysis. Nontransgenic and single transgenic controls (pools of 50 fish per genotype) were prepared as above and used for gating and compensation. A total of three larval peptidomic samples were collected on separate days, all within 1 week of each other.

For adult samples, procedures were largely similar. Male and female Tg(neurod1:RFP); TgBAC(cldn15la:GFP) adults were euthanized at 18 months post fertilization and their intestines were dissected, cut along the longitudinal axis to open the gut, and placed in chilled epithelial buffer with 10 μM ROCK inhibitor (5.6 mM Na₂HPO₄, 8 mM KH₂HPO_4, 96.2 mM NaCl, 1.6 mM KCl, 43.4 mM sucrose, 54.9 mM d-Sorbitol, 10 μM ROCK inhibitor). Samples rocked gently at 4°C for 15 min before being transferred to fresh epithelial buffer with ROCK inhibitor. Samples were shaken at roughly 150 shakes per minute to remove mucus and fecal debris. Intestines were then transferred to gentle MACS C tubes with 2 mL of disassociation buffer, as above. Intestines were digested with 3 rounds of protocol C_01 with 10 min of rocking at 4°C after each round. Identical to above, 10 mL of FACS media was then added, and contents were passed through a 30 micrometer strainer into a fresh 50 mL conical tube. The strainer was then washed with an additional 10 mL of FACS media and the contents were spun down, decanted, and resuspended in 1 mL of FACs media for immediate sorting. neurod1:RFP; cldn15la:GFP double positive, 7AAD negative cells from each of nine samples run in parallel were sorted into a single 1.5 mL LoBind tube (Sigma Z666505) with 500 μL of 6M guanidine hydrochloride for a total of roughly 50,000 cells. This constituted a single sample for downstream peptidomic analysis. Nontransgenic and single transgenic controls (1 dissected intestine per genotype) were prepared as above and used for gating and compensation. A total of three adult peptidomic samples were collected on separate days, all within 1 week of each other. All samples were stored at −80°C until they were transferred to the Duke Proteomics and Metabolomics Core Facility and processed for downstream peptidomics as described below.

LC–MS/MS analysis of peptidome samples

Samples sorted into 500 μL of 6M guanidine hydrochloride were subjected to five rounds of 30-s bath sonication (Branson) at 30% power with cooling in between. Samples were then diluted with 500 μL of 0.1% formic acid and centrifuged at 12,000 rpm to remove cellular debris. Lysates were loaded directly onto a 10 mg Oasis HLB SPE cartridge (Waters), washed with 2 × 1 mL of 2% acetonitrile/0.1% formic acid, and eluted with 250 μL of 50% acetonitrile/0.1% formic acid. Eluents were speed-vacuumed to dryness. Samples were then resuspended in 50 μL of 50 mM ammonium bicarbonate with 10 mM dithiolthreitol and heated at 70°C for 30 min, alkylated with 25 mM iodoacetamide for 30 min at room temperature and then spiked with 2 fmol/μL of pre-digested yeast alcohol dehydrogenase (Waters MassPrep standard).

LC/MS/MS was performed using an EvoSep One UPLC coupled to a Thermo Orbitrap Astral high-resolution accurate mass tandem mass spectrometer (Thermo). Briefly, each sample loaded EvoTip was eluted onto a 1.5 µm EvoSep 150 μm ID × 15 cm performance (EvoSep) column using the SPD30 gradient at 55°C. Data collection on the Orbitrap Astral mass spectrometer was performed in a data-dependent acquisition mode of acquisition with a r = 120,000 (@ m/z 200) full MS scan from m/z 300–2,000 in the OT with a target AGC value of 300% and max accumulation time of 50 ms. Data-dependent MS/MS scans in the Astral were performed on charge states 2–7 from m/z 110–2,000 at a target AGC value of 200% and max accumulation time of 20 ms. HCD collision energy setting of 30% was used for all MS2 scans. Data were imported into PEAKS studio. The software was set to 10 ppm mass accuracy on MS1 and 0.02 Da on MS2 data with no enzyme rules selected. De novo assisted searching was allowed. The MS/MS data was searched against a custom Danio rerio database along with a common contaminant/spiked protein database (bovine albumin, bovine casein, yeast ADH, etc.), and an equal number of reversed-sequence “decoys” for false discovery rate determination. Database search parameters included fixed modification on Cys (carbamidomethyl), variable modification on Met (oxidation), C-terminal amide, and N-terminal acetylation and pyroglutamate. We queried our raw data against a custom database of zebrafish proteins from TrEMBL [193] and Ensembl [97]. To account for genomic polymorphisms within the Ekwill zebrafish strain used here, we supplemented the combined TrEMBL/Ensembl database with protein sequences predicted with customProDB [194] from genetic variants observed in previous RNA sequencing from the same strain [195]. Spectral annotation was set at a maximum 1% peptide false discovery rate based on q-value calculations. Manual verification of spectra of interest were confirmed using targeted extraction/selected ion chromatograms with Skyline (University of Washington, MacCoss Laboratory) [196] including considerations for signal to noise, peak shape, mass errors across isotopologues, and retention time relative to database identifications. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD058654 [197]. In addition, we have shared screenshots of Skyline data for an example of a manually reviewed peptide in S5C and S5D Fig.

Ghrelin acylation search

To identify acylated ghrelin peptides, we performed a PEAKS search on a custom database just containing the F1QKX9 ghrelin sequences from Uniprot and selected octanoylation and decanoylation as variable posttranslational modifications in addition to the modifications in our original search. Spectral matches and chromatogram peaks of the identified peptides are included in the supplemental materials (S5A and S5B Fig).

Zebrafish secretin peptide search

As the putative zebrafish secretin homolog we identified (si:zfos-2372e4.1) did not have a TrEMBL entry, it was not listed in our reference dataset, and, as such, was not identified in our initial annotation of our peptidomics dataset. We first ran a preliminary search of the data with a custom dataset that just contained the protein sequences available on Ensembl for si:zfos-2372e4.1. We saw several hits, indicating peptides from that gene were indeed present in our sample. As searching with a drastically smaller reference dataset can inflate the possibility of identifying significant peptides, we added those two si:zfos-2372e4.1 protein entries to our pre-existing custom zebrafish reference dataset, and re-ran the larger search. In our final annotation (Fig 2F), we only included those peptides that met significance in this more stringent second search.

Protein alignments

Protein sequences for Danio rerio, Carassius auratus, Ictalurus punctatus, Salmo salar, Oncorhynchus mykiss, Aquarana catesbeiana, Gallus gallus, Rattus norvegicus, Mus musculus, Homo sapiens, Cyprinus carpio carpio, Tetraodon nigroviridis, Xenopus tropicalis, Gasterosteus aculeatus aculeatus, Dicentrarchus labrax, Myripristis murdjan, Clupea harengus, Rana temporaria, Oreochromis niloticus were downloaded from UniProt [193]. When more than one sequence was available, the sequence designated as canonical in UniProt was used. If there were multiple relevant entries, a representative one was used for ease of visualization. Sequences were aligned using MUSCLE [198] and visualized with JalView [199]. Percent identity settings were used to color residues.

For tachykinin analysis (S15 Fig), Molecular Evolutionary Genetics Analysis (MEGA) version 12 [200] was used to align protein sequences (method: MUSCLE [198]) and generate a phylogenetic tree shown in S15 Fig (method: Neighbor-Joining [201] with bootstrapping).