Gene regulatory patterning codes in early cell fate specification of the C. elegans embryo

  1. Alison G Cole
  2. Tamar Hashimshony
  3. Zhuo Du
  4. Itai Yanai  Is a corresponding author
  1. Department of Molecular Evolution and Development, University of Vienna, Austria
  2. University of Vienna, Austria
  3. Department of Biology, Technion – Israel Institute of Technology, Israel
  4. State Key Laboratory of Molecular Developmental Biology, Institute of Genetics and Developmental Biology, Chinese Academy of Sciences, China
  5. Institute for Computational Medicine, NYU School of Medicine, United States


Pattern formation originates during embryogenesis by a series of symmetry-breaking steps throughout an expanding cell lineage. In Drosophila, classic work has shown that segmentation in the embryo is established by morphogens within a syncytium, and the subsequent action of the gap, pair-rule, and segment polarity genes. This classic model however does not translate directly to species that lack a syncytium – such as Caenorhabditis elegans – where cell fate is specified by cell-autonomous cell lineage programs and their inter-signaling. Previous single-cell RNA-Seq studies in C. elegans have analyzed cells from a mixed suspension of cells from many embryos to study late differentiation stages, or individual early stage embryos to study early gene expression in the embryo. To study the intermediate stages of early and late gastrulation (28- to 102-cells stages) missed by these approaches, here we determine the transcriptomes of the 1- to 102-cell stage to identify 119 embryonic cell states during cell fate specification, including ‘equivalence-group’ cell identities. We find that gene expression programs are modular according to the sub-cell lineages, each establishing a set of stripes by combinations of transcription factor gene expression across the anterior-posterior axis. In particular, expression of the homeodomain genes establishes a comprehensive lineage-specific positioning system throughout the embryo beginning at the 28-cell stage. Moreover, we find that genes that segment the entire embryo in Drosophila have orthologs in C. elegans that exhibit sub-lineage-specific expression. These results suggest that the C. elegans embryo is patterned by a juxtaposition of distinct lineage-specific gene regulatory programs each with a unique encoding of cell location and fate. This use of homologous gene regulatory patterning codes suggests a deep homology of cell fate specification programs across diverse modes of development.

eLife assessment

This valuable work fills a gap in the mapping of gene expression patterns in the early embryo of C. elegans. The presented data are solid and provides a resource for future analysis.


Developmental toolkit genes constitute the genomic repertoire that underlies the pathways patterning the organism (Carroll, 2008). These genes are generally involved in signaling transduction modules and transcription factors (TFs) (Degnan et al., 2009; Gerhart, 1999). Their assembly into specific temporal and spatial gene regulatory pathways leads to the specification of cell fates in the embryo. For example, the Caenorhabditis elegans endoderm is specified by a cascade of TFs operating sequentially (Maduro and Rothman, 2002). Identifying the principles of gene regulation has been a central endeavor in developmental biology.

John Sulston and colleagues elucidated the 1340 cells generated throughout the embryogenesis of the nematode C. elegans, leading to a hatched worm with 558 cells, each with a described name, lineage, and fate (Sulston et al., 1983). The cell lineage is delineated into a germ-line lineage – P – and five ‘founder’ somatic cell lineages – AB, MS, E, C, and D – that each have characteristic cell cycle timings (Deppe et al., 1978) and cell fates (Figure 1a). All founder cell lineages are present by the 28-cell stage, when gastrulation initiates, and by the 102-cell stage the progeny of most of the cells have a single fate.

Single-cell transcriptomics in the early C. elegans embryo.

(a) The embryonic C. elegans cell lineage as deciphered by Sulston et al., 1983. The number of cells present at each stage is indicated on the left and schematic images of the embryo are shown for several stages to the right with cells colored by founder cell lineages AB, MS, E, C, D, and P4. The fate of the cells at the last stage is also indicated according to the legend at the foot of the lineage. The number of embryos examined in this study is indicated on the right. Colored bars correspond to the windows in which embryos were considered to be at the same stage. (b) A t-distributed stochastic neighbor embedding (tSNE) map of the cells isolated in this study. Circles indicate the individual cells, colored by the stage of development of the embryo from which they were manually collected. (c–d) Gene expression for the indicated lineage and fate marker genes on the tSNE map shown in (b).

The invariant cell lineage of the C. elegans embryo provides an opportunity to systematically study developmental gene pathways. Overall, there are 934 annotated TFs in C. elegans, with 40% having experimentally studied DNA-binding specificities (Narasimhan et al., 2015). The overall set can be classified into families according to their DNA-binding domain, including the homeodomain, bHLH, bZIP, T-box, C2H2 zinc finger, and the very large nuclear hormone receptor domain family. While mapping the location of signaling pathways is difficult from gene expression data, effects of signaling on the transcriptional output are detectable and consequently studying TF expression provides an entry point into gene expression pathway characterization.

Here, we use single-cell RNA-Seq to identify the transcriptomes of each of the cells up to the 102-cell stage. We find that homeodomain genes are expressed in stripes along the anterior-posterior axis, as early as the 28-cell stage. Interestingly, each founder cell lineage – AB, MS, C, and E – establishes its own regionalization code. These Drosophila-like stripe patterns suggest a deep homology in cell fate specification between the embryogenesis of non-segmented and segmented animals, despite differences in syncytium-based and cell-cleavage-based modes of development.


Previous studies analyzed cells from a mixed suspension of cells from different C. elegans embryos (Packer et al., 2019). However, this approach does not ensure sampling of each of the individual cells of the early stages, where it is difficult to dissociate the cells. A complementary approach is to manually collect cells by mouth pipette thus ensuring that all cells are collected. Other works employed this approach to study the early stages, however only arrived at the 15-cell stage (Hashimshony et al., 2016; Tintori et al., 2016). Up to this stage, few new TFs are induced and the transcriptome is dominated by the maternal deposit. As we were interested in studying early specification events in the embryo, we thus manually isolated individual cells from dissociated embryos and processed them for scRNA-Seq (Figure 1b). Overall, we studied 840 cells from 38 embryos up to the 102-cell stage, collecting all or most cells of each embryo (Figure 1b and Supplementary file 1).

Analyzing the transcriptomes, we found that embryo-to-embryo variation could be efficiently normalized by standardizing each gene’s expression across all cells collected from the same embryo (see Methods), and this is another benefit of the manual collection approach. A dimensional reduction map of the examined cells (Figure 1b) reveals the unfolding of development with cells from the early stages occupying the center while cells from later stages occupy the periphery. We found that cells followed developmental trajectories according to their founder cell origin, each identified by the expression of genes known to be expressed in a lineage-specific manner: ceh-51 (MS), elt-7 (E), pal-1 (C,D and P), pes-1 (D), and nos-2 (P) (Figure 1c; Broitman-Maduro et al., 2009; Hunter and Kenyon, 1996; Lee et al., 2017; Maduro and Rothman, 2002). As also observed by other studies (Packer et al., 2019), the founder cell lineage trajectories are oriented according to cell fates, with similar cell fates converging between lineages (Figure 1b and d): the cells of the ecto-mesoderm producing C lineage formed a bi-lobed cloud with the muscle part adjacent to that of the MS and D lineages. Similarly, MS cells that will form the pharynx orient toward the pharynx producing portion of the AB lineage.

In order to infer the precise identity of each cell, we first organized the embryos into eight developmental stages: the 1-, 2-, 4-, 8-, 15-, 28-, 51-, and 102-cell stages (Figure 1a). For each stage, we studied cells from multiple embryos and identified clusters of cells according to differential gene expression. Figure 2a demonstrates this analysis for the 15-cell stage and Figure 2—figure supplement 1 for all other stages. The identified clusters contain cells from each of the individual embryos, thus confirming our underlying assumption that embryo to embryo gene expression variation is minimal relative to the coherence of the cell states within an embryo. Finally, we inferred the cell identity of each cluster with reference to known gene markers from the literature (Supplementary files 2 and 3). Figure 2b illustrates our approach for the cells of the 28-cell stage (see Figure 2—figure supplements 27 for all other stages). We required a new marker to distinguish some of the cell identities. Studying F19F10.1 by single-molecule fluorescent in situ hybridization (FISH), we found that this previously uncharacterized ets-factor is expressed in the posterior daughter cell through 2-cell cycle divisions (Figure 2c). We further tested for the robustness of the cell identity assignments and ensured that each cell is most similar to its assigned identity (computed with itself excluded; Figure 2—figure supplements 27, bottom of each subfigure). We identified 5433 genes as being differentially expressed within a stage, collectively across all of the stages (see Methods). Of these, 395 are TFs during this period of cell specification (Supplementary file 6).

Figure 2 with 9 supplements see all
Inferring the transcriptomes of cell states throughout specification in the C. elegans embryo.

(a) t-Distributed stochastic neighbor embedding (tSNE) clusters of cells from 15-cell stage embryos, shown in terms of their embryo of origin (top), cell clustering (middle), and inferred cell identity (bottom). (b) Gene expression for the indicated genes across the assigned cell states of the 28-cell stage. Rows correspond to genes and each bar within a cell state indicates the expression level of a sample. Black boxes indicate differential gene expression. Colors in the bottom row indicate the embryo of origin. (c) Double single-molecule RNA in situ hybridization showing spatial identity of F19F10.1 expressing nuclei (yellow). The positions of all nuclei are shown with DAPI staining (blue), and the embryos were oriented with the posterior landmark gene pal-1 (purple) (Hunter and Kenyon, 1996). Cell identities were manually annotated based on previous descriptions (Supplementary file 4). (d) Average expression of cluster identity genes through the first 8 cell divisions of embryonic development. Each bar represents the inferred transcriptome of a cell state and shows the standardized expression for the indicated genes, using the same color bar as in (b). Equivalent transcriptomes derived from sister cells are indicated by blue circles. Lineage symmetry gives rise to left/right equivalence groups indicated with black boxes. (e) Expression of ceh-43 according to lineage of the inferred transcriptomes (top) and a transcriptional reporter strain (bottom), where highest expression is shown in red and no detected expression is black (see also Figure 2—figure supplement 9).

For some cell identities within a stage we were not able to distinguish distinct transcriptomes, and we thus denoted these as composing ‘equivalence groups’. These may either be attributed to daughter cells whose transcriptomes have yet to diverge, such as cells Ca and Cp in the 15-cell state, or to sets of cousins whose sub-lineages form repeating units, such as cells ABplp and ABprp (also of the 15-cell stage) (Sulston et al., 1983). In the first case, this indicates that sister cells are transcriptomically very similar shortly after division. Altogether we describe 119-cell identities with distinct transcriptome profiles, which by stage, are as follows: 1 (1-cell), 2 (2-cell), 4 (4-cell), 6 (8-cell), 11 (15-cell), 20 (28-cell), 38 (51-cell), and 37 (102-cell), with 64 of these representing equivalence groups of at least two cell states (Figure 2d, Supplementary file 5, and Figure 2—figure supplement 8). For each cell state that we assigned, we have an average of seven replicate cells (Supplementary file 4). We validated our annotations by imaging GFP reporters and found good correspondence, as illustrated for ceh-43, dmd-4, and unc-30 (Figure 2e and Figure 2—figure supplement 9). The delay between the GFP expression and the mRNA detected here is expected as previously described (Murray et al., 2012).

We next asked if the genes that are members of the same gene family have similar spatial and temporal expression arrangements. To test for this mode of expression, we considered groups of genes with the same DNA binding domain, as annotated by the PFAM database (Finn et al., 2016). For each cell, we queried for enrichments or depletions of genes of specific domain families, among its expressed TFs. As an example, the ‘ABplppp’ cell differentially expresses 14 TFs according to our analysis, including seven homeobox domain (HD) TFs, though overall HDs only account for 11% of the TFs (Figure 3a). We used the hypergeometric distribution to compute a significance of p=0.00015 for this overlap, given the overall number of TFs and the number of HD and ‘ABplppp’ TFs. Conversely, only two of the 69 TFs expressed by the Eala cell are HDs, a significant depletion (p=0.01, hypergeometric distribution). Figure 3b shows the overall pattern of enrichments and depletions of HDs across all of the examined cells, indicating enrichment of HDs in the 51-cell stage, in particular in the ABp lineage, and depletion in the E lineage. For the GATA type zinc factors we found enrichment in the E lineage, as well as some of the AB cells (Figure 3c).

Transcription factor (TF) gene families have spatiotemporal specificities.

(a) Detecting enrichment and depletion of genes of the same gene family, among the TFs expressed by a cell, using the hypergeometric distribution. (b–c) For the homeobox domain (b) and GATA zinc finger domain (c) gene families, the enrichments and depletions are indicated for each cell of the lineage, plotted as in Figure 2d. (d) TF family stage and lineage enrichments. For each cell (column), the enrichment and depletion for the genes of a TF family (row) is indicated (similar to b, c). The top bars indicate the lineage and stage of the cell. (e) For the indicated TF family the significance for enrichments is shown using the t-test between the values shown in (d) for the noted stage and lineage.

Examining the pattern of such biases, we found that TF families showed significant enrichments across lineages and developmental stages (Figure 3d–e). Four TF gene families are enriched in the AB lineage: the T-box domain, the helix-loop-helix (HLH) DNA-binding domain, the HD domain, and the high mobility group (HMG) box domain (Figure 3d). Each of these families also has enrichments for specific AB stages, from early expression (T-box and HLH) to later expression (HD and HMG).

From Figure 3 analysis we concluded that the homeodomain (HD) TF family in particular exhibited lineage-specific expression. Studying this in more detail, we observed that in the 28-cell stage, each of the 16 AB cells contains expression of at least one homeodomain gene, with 8 distinct cell state patterns (Figure 4a). This is mostly also apparent in the 51- and 102-cell stages. The expression patterns of all homeodomain genes across the lineage are not mono-phyletic. For example, ceh-43 is expressed in nine of the sixteen AB cells at the 28-cell stage in a complex pattern (Figure 4a). Thus, continuous symmetry breaking (Zacharias and Murray, 2016) does not appear to be a mechanism by which differential expression is generally established.

Homeodomain genes are expressed in lineage-specific stripes across the anterior-posterior axis.

(a) Lineage expression patterns for the indicated genes in the 28-, 51-, and 102-cell stage. Thick and thin lines indicate binarized ‘on’ and ‘off’ expressions, respectively. (b) Three-dimensional expression patterns separated by lineage (AB/MS/C) for the genes indicated in (a).Circles represent cells in the embryo at the particular stage examined. Expression of each gene is indicated by a circle of a different size and color. Light gray cells indicate cells in other lineages. (c) Clustering of cells according to gene expression of the indicated lineage and stage. Colors distinguish K-means clustering.

We asked if the relative spatial location of the cells within the embryo can provide insight into the observed homeodomain expression patterns. For this analysis, we mapped the binarized expression of genes onto the known three-dimensional locations of cells in the embryo. Studying the 28-cell stage, we found that the homeodomain genes expressed in each lineage revealed dorsal-ventral stripes along the anterior-posterior axis (Figure 4b). For example, the ceh-43-expressing cells, deriving from distinct AB sub-lineages, are physically proximate. Clustering the AB lineage cells into eight groups according to their binarized expression, we mapped the gene expression combinations of each stripe. One stripe, for example, includes the expression of unc-30, ceh-36, and ceh-13; the adjacent stripe has expression of unc-30 and ceh-36 (Figure 4b). At the 51-cell stage, 4 new homeodomain genes are expressed in the AB lineage, and again position along the anterior-posterior axis correlates with gene expression. In the 102-cell stage, expression becomes more complicated as the AB lineage forms an external shell around the gastrulating non-AB cells. Generalizing this approach, we also found distinct positional systems for the MS and C lineages (Figure 4b and c). For the MS lineage, ceh-32 and ceh-51 are in the anterior and posterior, respectively, while for the C lineage vab-7 (even-skipped) and hmbx-1 are in the anterior and posterior, respectively. We thus provide evidence that distinct regionalization codes occur across the embryonic sub-lineages.

To compare the patterning code elucidated for C. elegans embryogenesis to the well-studied patterning pathway in Drosophila, we also studied the expression of C. elegans orthologous genes. Drosophila development proceeds through a series of molecular events which partition the syncytial embryo along the anterior-posterior axis (Wolpert et al., 2019). We found that the C. elegans genes such as nos-1,2 (nanos) and puf genes (pumilio) which are maternal genes in both systems are restricted to the primary germ cell lineage in C. elegans (Figure 5). Moreover, gap and pair-rule genes, such as pal-1 (caudal) and cwn-1 (wg), show C lineage and sub-lineage restriction, respectively (Figure 5). Thus, it is evident that while Drosophila patterning follows in a lineage-independent manner, in C. elegans, lineage dependency is coincident with specific expression of patterning genes across the founder cell lineages.

Expression of the orthologs of Drosophila specification genes in C. elegans.

Lineage expression patterns in the 28-cell stage for the indicated genes whose orthologs function as maternal, gap, pair-rule, and segment polarity genes in Drosophila. Thick and thin lines indicate binarized ‘on’ and ‘off’ differential expressions, respectively.


Collectively, our analysis reveals that during embryogenesis, in between the early signaling period – involving Notch (Priess, 2005) and Wnt (Rocheleau et al., 1997; Thorpe et al., 1997) signaling – and the later cell fate differentiation period – involving myogenic factors such as myoD/hlh-1 (Chen et al., 1992) and GATA/elt-2 (Fukushige et al., 1998) – there is a period of lineage-specific patterning. Three of the homeodomain genes whose expression we have described – ceh-13, php-3, and nob-1 (Figure 4) – are members of the HOX sub-family, known to provide regionalization during embryogenesis in bilaterians (Carroll et al., 2004). Indeed, we observed that expression of these HOX genes transcends the founder lineages and patterns the posterior region in multiple lineages (Figure 4). However, we find 51 other homeodomain genes that likely pattern the lineages and set up locations in the embryo. This is a large genetic repertoire relative to Drosophila (97 genes in C. elegans, relative to 104 in Drosophila) despite the higher complexity of the latter in terms of cell types and total number of cells. Thus, in species where regionalization is under lineage control, as opposed to morphogen gradients, there is apparently a greater necessity for distinct positional systems across distinct sub-lineages – here the founder cell lineages. We find lineage-specific gene expression stripes (Figure 4), suggesting that such a developmental patterning mechanism may not be restricted to segmented organisms but is rather a universal feature of animal embryogenesis. The centrality of homeodomain genes to this system highlights the adaptability of this TF family to both lineage-dependent and position-dependent functions. It will be interesting to further compare other species at the single-cell level in order to trace the evolutionary origin of gene expression developmental patterning codes.


Cell isolations

Embryos were isolated from adult worms and removed from their egg shells as described in Edgar and Goldstein, 2012. Individual embryos were manually dissociated by gentle pipetting through progressively smaller glass-bore pipettes in calcium/magnesium-free EGM. Cell collection was limited to a 30 min window from the beginning of the dissociation process. For embryos with more than 28 cells, collection was performed as a team. Cell losses during dissociation were noted and single dissociated cells from an individual embryo were counted, collected as previously described (Hashimshony et al., 2012), and stored at –80°C until further processing.

Single-cell RNA-Seq

Individual embryo, multiplexed single-cell RNA-Seq libraries were produced following the CEL-Seq2 protocol (Hashimshony et al., 2016). Briefly, individually harvested cells were barcoded with a reverse transcription reaction, and RNA molecules were tagged with unique molecular identifiers (UMIs) (Kivioja et al., 2011). All barcoded cells from a single embryo were then combined into a single IVT amplification reaction which was used to generate an Illumina library for sequencing. Multiple sequencing libraries with compatible Illumina sequencing indexes were sequenced together on an Illumina HiSeq 2500 using standard paired-end sequencing protocols. For Read 1, used to determine the barcode, the first 15 bp were sequenced and for Read 2, used to determine the identity of the transcript, the first 35 bp were sequenced. The CEL-Seq2 pipeline is available at GitHub (copy archived at YanaiLab, 2017). Mapping of the reads used BWA33, version 0.6.1, against the C. elegans WBCel215 genome (bwa aln -n 0.04 -o 1 -e -1 -d 16 -i 5 -k 2 -M 3 -O 11 -E 4). Read counting used htseq-count version 0.5.3p1 defaults, against WS230 annotation exons.

Initial processing of the CEL-Seq2 data

Gene expression values were calculated as UMI-corrected reads normalized to the number of transcripts per 50,000; only cells with more than 10,000 mapped reads were included in the analysis. In order to compare gene expression patterns between embryos, gene expression was first standardized among the cells of an individual embryo (such that the mean expression for each gene across the cells of the embryo is 0 and the standard deviation is 1), and these standardized expression profiles were used for comparisons between embryos.

Cell clustering

t-Distributed stochastic neighbor embedding (tSNE) (van der Maaten and Hinton, 2008) was performed on full dataset using the top 1000 most variable genes (according to the Fano factor), with 20 initial dimensions and perplexity set at 50. For reproducibility, a non-random initiation of the objective function was set (

Cell identity inference

For each of the eight examined stages – 1-cell, 2-cell, 4-cell, 8-cell, 15-cell, 28-cell, 51-cell, and 102-cell – the isolated cells were first clustered according to their position on a tSNE plot using genes of known expression patterns (Supplementary file 2). We next manually and iteratively accepted clusters based upon their homogenous embryo composition, by removing the accepted cluster and re-clustering the remaining cells. These clusterings were based upon the restricted set of genes with known cell identity (Supplementary file 2). Cell state was then annotated to the clusters by comparing the average gene expression of the cluster with a collated database of known gene expression patterns (Supplementary file 3). The robustness of the clustering was tested by only retaining cells whose expression is most similar to the mean expression of each cluster, computed without it (Figure 2—figure supplements 27). The cell state transcriptome was then computed as the average expression of all members, together with standard deviation (Supplementary file 6).

FISH analysis

Custom Stellaris FISH probe sets were designed against the entire coding sequence of F19F10.1 (25 non-overlapping probes labeled with CAL Fluor Orange 560 Dye) and pal-1 (25 non-overlapping probes labeled with C3-Fluorescein Dye), using the Stellaris FISH Probe Designer (Biosearch Technologies, Inc, Petaluma, CA, USA). FISH was performed according to the protocol available on the Stellaris website for fixation of embryonic C. elegans material, with the following modification: all steps were performed in 1.5 mL Eppendorf tubes. Hybridization of two non-overlapping probe sets was performed using 125 nM of each probe set contemporaneously. Within 48 hr from completing the hybridization protocol, embryos were mounted on glass slides under #0 coverglass (85–115 µM thickness), supported on each corner by small amounts of plasticine to avoid crushing the embryos. Slides were imaged immediately after mounting with a Zeiss LSM 510 confocal microscope.

Lineage tracing and single-cell fluorescence quantification

Embryo mounting and live imaging were performed as previously described. Briefly, 2- to 4-cell stage embryos were mounted between two cover slips in M9 buffer containing 20 μm microspheres (Polyscience). Dual-color 3D time-lapse imaging was performed using a Spinning disc confocal microscopy. Images were acquired with 30 focal planes at 1 μm Z resolution for every 75 s during a 6 hr duration of embryogenesis. Ubiquitously expressed mCherry was used for automated cell identification and lineage tracing using StarryNite program followed by systematic manual correction of errors (Santella et al., 2014). Protein fusion GFP was used to quantify expression level of corresponding gene in each cell using a previous established method (Murray et al., 2008). GFP intensity was measured by calculating the average GFP intensity for each cell and then with the local background intensity subtracted. GFP intensity at different time point for a same cell was averaged to quantify expression level of a gene in a cell. These data are further described in the recently published atlas (Ma et al., 2021).

Differential gene expression analysis

For each gene, at each stage, we identified differential gene expression using the following approach. We compared the standardized gene expression values across the cell states of the stage, and identified the two groups of cell states that correspond to the most significantly different groups, assayed by a t-test p-value. In order to establish the significance, we then asked if that p-value is better than that expected by shuffling the expression values across states. To find the two groupings with the best p-value, we began by computing a t-test between the expression values for each cell state and the rest of the expression levels of the remaining states. The cell state with the best p-value was used to seed a group and we next computed the p-value of a super-group composed of this cell state in turn with each other cell state, comparing with the expression values of the remaining cell states. If one of these super-groups had a better p-value than the original best p-value state alone, we selected it and continued this process, iteratively adding to the super-group until the p-value can no longer improve. This process results in a 1-cell state or a super-group of cell states with expression and a group without expression. Each gene will thus have a binary profile and an associated p-value. To select the genes with sufficiently significant p-values we repeated the procedure on 10,000 shuffled sample to cell state assignments. Sorting the p-values from the shufflings, the p-value at 1% of this list was used as the threshold, for a false discovery rate of 1%. We excluded differential expression of genes expressed in more than 50% of the cells of a given stage, and also restricted differential expression to those genes whose maximum expression is at least three quarters of a log10 TPM unit greater than the minimum expression at the given stage.

Enrichments and depletions of TF families in cells

For each cell, we computed both the enrichment and depletion of the set of genes in a given TF family among its expressed TFs. The enrichment and depletion is computed as the cumulative hypergeometric p-value with and without subtracted from unity, respectively. The smallest number between these two tests distinguishes an enrichment from a depletion (see Figure 3a), and we multiplied the –log10 p-value by 1 and –1, to distinguish these in the plots shown in Figure 3b, c. To compute the significance of enrichments for particular lineages and stages, we used the t-test on the –log10 p-values of the enrichment values.

Materials availability statement

The complete dataset has been deposited to the NCBI GEO database GSE83523. The authors declare no competing financial interests. Correspondence and requests for materials should be addressed to IY (

Data availability

The complete dataset has been deposited to the NCBI GEO database GSE83523.

The following data sets were generated
    1. Cole AG
    2. Hashimshony T
    3. Yanai I
    (2023) NCBI Gene Expression Omnibus
    ID GSE83523. Transcriptome characterization of embryogenesis at a single cell resolution [C. elegans].


  1. Book
    1. Carroll SB
    2. Grenier JK
    3. Weatherbee SD
    From DNA to Diversity: Molecular Genetics and the Evolution of Animal Design
    1. Fukushige T
    2. Hawkins MG
    3. McGhee JD
    The GATA-factor elt-2 is essential for formation of the Caenorhabditis elegans intestine
    Developmental Biology 198:286–302.
    1. van der Maaten L
    2. Hinton G
    Visualizing Data using t-SNE
    Journal of Machine Learning Research: JMLR 9:2579–2605.

Peer review

Reviewer #1 (Public Review):

C. elegans is a pre-eminent model for developmental genetics, and its invariant lineage makes it possible in theory to define molecular features such as gene expression comprehensively and at single cell resolution across the organism.

Previously published single-cell RNA-seq studies have mapped gene expression across the lineage through the 16-cell stage (Tintori et al 2017, Hashimshony et al 2016), and at later stages (Packer et al 2019, with good coverage starting at the 100-cell stage and some coverage at the ~50-cell stage). This left the critical period around gastrulation (~28-cell and ~50-cell) without comprehensive transcriptome data. This study covers this gap with a heroic effort involving the manual isolation and analysis of over 800 cells from embryos of known stage, combined with painstaking curation using known markers from small scale studies and larger imaging-based expression atlases. Importantly, the dataset overlaps at early and late stages with data prior studies.

The data quality and overlap with Tintori and Packer datasets both appear high, but to make this inference required additional analysis from Supplemental Table 6 by this reviewer as it is not explored or described in the manuscript. Analyses demonstrating continuity with these datasets would greatly increase the value of the resource.

The authors show that specific lineages and stages preferentially express TFs with different classes of DNA binding domains. This extends previous work implicating homeodomains as preferentially involved in nervous system patterning and as enriched in neural and muscle progenitors in mid-stage embryos.

They also show that C. elegans homologs of Drosophila early embryonic regulators (which function based on spatial position in that system) tend to also be patterned in early C. elegans embryos, but with lineage-specific patterns. This conserved use of regulators would be fairly remarkable given the dramatically different developmental modes in these two species, although this observation is not backed up by quantitative analyses.

Finally, there is an argument that combinations of TFs expressed in lineage-specific patterns give rise to "stripe" patterns. This section is also not based on statistical analyses but suggests the possibility that lineage and positional regulation may be more convoluted than was previously thought.

Reviewer #2 (Public Review):

The C. elegans embryo has been model system of study for more than 30 years because of the ease of doing forward and reverse genetics, coupled with its nearly invariant lineage which allows a description of development at high resolution. 4D time lapse imaging coupled with spatially resolved gene expression has enabled identification of transcriptional signatures of cells in space and time, and in the past decade this has been advanced with single-cell transcriptomics methods, using individually isolated embryonic cells (which can retain their identity) or by deconvolving complex mixtures of early cells. Recent work using these methods has resolved spatiotemporal expression patterns for many genes, defining cells up to gastrulation stage, but then changing to more tissue-specific patterns during morphogenesis. A key paradigm of specification in C. elegans and other systems is that early maternal factors initiate or restrict patterns of transcription factor expression from the zygotic genome. Combinatorial expression patterns and some symmetries broken by autonomous or extrinsic cell inductions ultimately program lineages towards their fates. To date, only simple networks have been elucidated, as the increasing complexity of these networks and the high level of redundancy has made functional dissection of such pathways difficult. Hence, almost all of the work in recent years has been descriptive.

In this work the authors fill a knowledge gap from the early embryo (~16 cells) to the ~100-cell stage and describe new patterns of gene expression. They reconcile their findings with that of others who have defined expression patterns using other methods, such as scRNA-Seq from complex mixtures of cells, and from transcription factor expression analyses. The resulting description of embryonic develop is the most precise to date, and offers a potentially useful resource for other researchers.

The authors attempt to use their results to find patterns of gene expression that could hint at phylogenetic conservation of specification mechanisms. They find some supporting evidence that expression of homeobox genes occurs in anterior-posterior stripes, which recalls the elaborate A/P patterning system elucidated in the Drosophila embryo, which belongs to the sister phylum Arthropoda in the Ecdysozoan clade of molting animals. It felt as if the authors chose the Hox genes they need to support this conclusion.

Some caveats exist to the work. The expression patterns seem to be well-validated, and following prior work from the Yanai group are likely to be strongly correlated with expression in living embryos. When cells are separated, they could lose some expression patterns that require cell-cell interactions, so it is expected that there might be a small minority of expression patterns that are more complex than what has been documented here.

A major caveat is the idea of the stripes of Hox expression. I just did not find these arguments to be compelling. Seeing these 'stripes' requires organizing the data in a way that maximizes their appearance, for one. Since there is not a lot of movement of cells away from their birth in the early embryo, the AB descendants are anterior to those of MS, anterior to those of E, anterior to those of C, D, and P4. Lineage-specific expression will just naturally fall into 'stripes'. Second, the conservation of Hox expression patterns typically comes with collinearity of the genes along the length of a chromosome (i.e. the so-called Hox clusters) with expression along the body axis, as well as posterior-to-anterior fate transformations when Hox specification is disrupted.

A minor note is the detection of an enrichment of GATA factors in the early E lineage. This has now been found to be a derived condition even within the genus see Broitman-Maduro et al. Development 149 (21): dev200984, as other species like C. angaria show only a simpler network of elt-3 -> elt-2. This suggests that many of the other patterns of gene expression, particularly in the early embryo, could be highly derived as well; some caution is warranted in generalizing the results as being conserved with arthropods as some of this could be convergent.

Given what the authors are proposing about Hox stripes, some omissions of prior work were surprising (or maybe I missed them). For example, a comprehensive study of Hox genes in C. elegans by Hench et al. (2015) (PLoS One 10(5): e0126947) evaluated all the homeobox genes and examined their genomic locations and expression patterns in the embryo at high spatiotemporal resolution. Work from the Hobert lab (Nature 2020, 584(7822):595-601) showed how homeobox codes specify classes of C. elegans neurons, and the Murray lab (PLoS Genet. 18(5):e1010187) examined Hox control of posterior lineage specification at high resolution, with functional assays.

The Discussion section of the paper is brief, consistent with the descriptive nature of the work overall, but it would have been nice to see the findings related to other published studies as indicated above.

Reviewer #3 (Public Review):

The authors claim that this dataset covers a timepoint of embryogenesis that is not well covered in the other published single cell datasets (Tintori et al 2016 and Packer et al 2019). The Tintori data indeed do not cover the 28-102-cell stages sufficiently, but it is unclear how the data presented here compare to the Packer et al data. It is true that the Packer et al data have fewer cells at earlier timepoints than at later ones, but given that they sequenced tens of thousands of cells, they report that they still have ~10,000 cells <210 min of embryogenesis. It seems that if the authors want to make any claims about how their data enables exploration of a stage that was previously not accessible, this would require a better comparison to the available data.

The authors provide thorough support for how they assigned cell identities in their data. It is surprising though that at the 102-cell stage they only identify 37 unique cell identities. They suggest that this is because there are many equivalence groups at this stage. However, I would strongly encourage the authors to perform a similar analysis or otherwise compare their obtained identities with the data from Packer et al. 2019. It seems possible that given the low number of cells in this dataset, the authors are missing certain identities and it would be important to know this.

The main analysis the authors perform is to look at expression patterns of various classes of TFs and ask whether they are enriched in particular lineages or at specific timepoints. This analysis is interesting but would be more informative if the authors provided in Figure 3d the numbers of each class of TFs. The authors then focus on the homeodomain class of TFs as they display interesting lineage-specific expression patterns, which when mapped on the embryo form stripes. The stripe pattern however is not that obvious, at least not as shown in Figure 4b. Perhaps separate embryo schematics showing the different TF expression patterns would show this more clearly. Moreover, given the relatively small number of cell identities found in this dataset (particularly at the 102-cell stage), a similar analysis using the Packer data would provide further support to these patterns. The localization of cells with shared expression patterns does show a stripe pattern at the 28-cell stage, but also not so clearly beyond this timepoint.

I am also unsure about the validity/value of the comparison of the stripes to Drosophila and the centrality of homeodomain TFs to anterior-posterior positional identity. First, it would be important to map other TFs, very likely there are several other TFs that correlate with positional identity. Also, even if the expression of the homeodomain TFs in C. elegans form stripes, there are still several cells within that stripe that do not express these TFs, it is thus unclear whether these TFs encode positional information or the identity of cells with different positions in the embryo.

Reviewer #4 (Public Review):

This is an admirable piece of work. The authors build on a previous dataset they assembled, but expand it to include all stages of early development in the nematode Caenorhabditis elegans. Cell collection was done manually, which is very impressive, and is clearly far better than pooled unidentified cells. I will not comment on the specific sequencing and analysis, since this is not my expertise, but will comment on the general conclusions and comparative framework in which the authors place their results.

While the Introduction and Discussion sections are actually fairly short, much of the presentation of the results is based on a certain comparative framework, which is explicitly a comparison between C. elegans and Drosophila melanogaster. This is an important perspective, but I feel the authors' interpretation is in some places exaggerated and in other places almost trivial.

Drosophila and C. elegans are two of the main models for developmental biology. However, it has been clear for over two decades that both species are highly derived and specialized and therefore, treating them as representative for their taxa is problematic. Much of the authors' discussion hinges on the question of comparing syncytial and lineage-dependent development. The syncytial early development of Drosophila is very specific and is clearly a recent innovation within a restricted group of flies. The canonical Drosophila segmentation cascade is mostly a novelty and most elements within the cascade are recent. Specifically, the expression of gap genes in regional stripes is not found very broadly. Conversely, the polarizing role of Caudal is very ancient and is probably found in all Bilateria. When making comparisons with a distantly related species, it is important to keep this in mind. Not as much is known about development of other nematodes, but the little that is known indicates that C. elegans is also unusual, and specifically the eutelic development (conserved cell lineages in development) is not found in all nematodes.

The authors suggest that regional expression of transcription factors in stripes is a conserved characteristic of development. This is true for Hox genes and has been known for decades. The regional expression they show for other genes is not convincing as "stripes". It is no surprise that developmental transcription factors are regionalized, but linking this to the stripes of Drosophila gap genes and even more so to Drosophila pair-rule and segment-polarity genes is a bit far-fetched. Yes, many genes are expressed in restricted domains along the A-P axis, but that is all that can be said based on the data. Calling them "Drosophila-like" is unfounded.

Author response

The following is the authors’ response to the original reviews.

Reviewer #1 (Recommendations For The Authors):

1. In general given several of the "equivalence groups" were distinguished from each other in Packer et al's annotation, can the authors comment more on why they aren't able to distinguish them? Are the markers listed for those cell states in Packer not expressed appropriately in these data? Or are they expressed but the states are not different enough to form discrete clusters? I suggest the possibility that the analysis choices of 20 "initial dimensions" or 1000 most variable genes filtered out some of these differences which may be encoded in later principle components, or that the use of t-SNE projection is not sufficient to resolve these distinct states.

1. I was a bit confused by the spatial gene expression analysis. Several distinct ideas appear to be posed in the text. These ideas aren't really supported by any quantitative analysis, just the visual patterns in Figure 4B/C which I'm not sure I always agree with.

For example, ceh-43 expression is mentioned as having "physically proximate" expression. But it is well established that different lineages form specific spatial territories (e.g. Schnabel et al 1997). Thus it seems logical that genes with specific lineage patterns will have specific spatial patterns as well. If the claim is that the observed patterns are more clustered along the A-P axis than expected by chance given their lineal complexity then I'm not sure this is shown. Maybe some comparison with control lineage patterns of similar complexity of non-TFs or non-HD TFs could get whether these genes specifically are more spatially patterned? Visually it looks to me like some patterns are more like "blobs" or even lateral or D-V specific patterns than they are like "stripes."

In addition there is a long history in the literature discussing the origin of position-specific patterns in C. elegans - most I'm aware of support the idea that positional information arises primarily from intrinsic lineage mechanisms (e.g. Cowing and Kenyon 1996). Perhaps the authors are making this same argument here, but if so this isn't clear from the text.

Or maybe the authors are trying to make the argument that combinations of TFs encode more precise position than individual TFs? This seems more likely to me from the images presented still not well-supported without quantitative or statistical analyses.

1. The comparison with Drosophila is interesting but also under-developed. I think all I would feel comfortable claiming from the data as shown is that genes that are spatially patterned in early fly development are also usually patterned in the C. elegans lineage. But to even say this is an enrichment over expectation would require more analysis.

Minor comments:

Methods: some statement about temperature control during cell isolation would be useful. In other words were embryos continuing to develop or put at low temperature such as in a cold room to prevent temporal differences between the first and last cells collected from a given embryo?

Current links to data at GEO are incorrect and link to Levin et al 2016 instead. I was not able to access the raw single cell data, just the processed data in Table S6.

The standardization of expression in embryos isn't well explained - would be good to expand a little on the types of batch effects being addressed and how this approach was chosen or a relevant citation.

Page 2: Including P0 and cell deaths there are 1,341 branches in the hermaphrodite lineage (2n-1 for 671 terminal cells including deaths).

-"as their each have" (grammar error)

-"very large nuclear hormone receptor domain" (add "family")

Page 3: As noted Packer et al largely missed cells prior to the 50-cell stage as described - but the reason for this is likely that the use of 10 micron filters or centrifugation to remove undissociated embryos also removes early stage cells.

-"few new expressions occur" (grammar). Also, in both Tintori and Hashimshony datasets there well over 1000 newly expressed genes detectable (see for example Sivaramakrishnan et al 2021 biorxiv).

Figure S1 would be easier to interpret with a legend explaining what fates are represented by each color

Some genes listed as markers in Figure S2 are not included in the marker table such as flh-3, oma-2, sma-9.

"New markers were required" - this is plural but only F19F10.1 is mentioned. Were other markers examined this way or should it be singular?

In Figure S2 the lower ("robustness") plots are nice but could be explained more clearly. What is the nature of the "cell similarity score"? How many (if any) cells were excluded due to not being most similar to their own cluster?

"transcriptomically very similar shortly after division" - can the authors comment on any information they have about how long after division the cells were collected?

GFP reporter lineaging - the methods are minimally described (what brand of microscope, which strains/transgene/CRISPR configurations etc). And data are not presented. If these embryos are all incorporated into Ma et al 2021, that is fine, but should be clearly cited. Otherwise it is important in my view to include some way to access the quantitative values from the lineaging and understand these details.

"as illustrated for ceh-43, dmd-4 and unc-30" - were there other examples as suggested from this wording? I'd also note that similar fluorescent reporter imaging data have been published previously for all three genes listed (Walton et al 2015 for UNC-30, Ma et al 2021 for DMD-4 and CEH-43 protein reporters, Murray et al 2012 for dmd-4 and ceh-43 promoter reporters).

Zacharias and Murray are cited as promoting "continuous symmetry breaking" but actually that review argued for a "non-monophyletic" architecture similar to that supported by the data .

The text and figure don't always agree. For example mec-3 expression is listed in the text as part of one of the stripes, but mec-3 is not labeled on the figures.

The stage of each embryo in figure 4B/C should be explicitly labeled (and maybe also given specific figure panel designations to clarify what statements in the text correspond to which figures).

In the discussion it is unclear what the numbers "97 to 104" refer to

The scRNA-seq reads were mapped to a relatively old genome build and annotation set (WS230) - thus current users may find discrepancies with current gene names in WormBase. Also, since the CEL-seq data are 3' biased, it is worth noting that Packer et al found that a substantial number of genes (~1000) in a slightly later annotation set (WS260) were undercounted (sometimes dramatically) with the similarly biased 10x data due to incomplete 3'UTR annotations. While I would be reluctant to ask for a requantification for the purposes of the manuscript given the challenges of repeating the various analyses, it is worth explicitly mentioning whether this was dealt with.

Reviewer #2 (Recommendations For The Authors):

The writing was otherwise good, at least to my eye, and the data was presented very well and made freely available to other researchers. I am not as well-versed in the statistical methods and will leave comments on these to a better-equipped reviewer(s).

Fig. 1 legend 'P' should be P4 (subscript 4).

p. 9 'ceh-51' should be italicized.Only one factor seems to have been confirmed by smFISH, F19E10.1. There are available reporters, did they show a similar pattern? From CGC website: RW12347F19F10.1(st12347[F19F10.1::TY1::EGFP::3xFLAG]) V endogenous tagged reporter; RW11620 unc-119(tm4063) III; stIs11620 [F19F10.1::H1-wCherry + unc-119(+)] array reporter.

Reviewer #3 (Recommendations For The Authors):

Typo: on page 11, where it says nanog it should read nanos.

Reviewer #4 (Recommendations For The Authors):

I found some sentences and paragraphs to be a bit unclear. There are no page or line numbers in the manuscript, so I point in the general direction, and hope the authors find what I am referring to.

  • 2nd paragraph of the Introduction - "their" should be "they", but the sentence as a whole is not clear.

  • 3rd para. of the Intro. - The last sentence of this paragraph doesn't make sense. Please rephrase and/or break up into shorter sentences.

  • 1st Para. of Results - "the maternal deposit" is not clear. Perhaps "maternally deposited transcripts" or something similar.

  • 1st Para. after Figure 3. The last sentence "Thus, continuous symmetry breaking..." is unclear. What is "continuous symmetry breaking"? Please define and expand.

  • Fig. 4 - the genes seem to be listed from posterior to anterior. The common way of presenting Hox gene lists and other regionally expressed genes is from anterior to posterior.

  • For the benefit of the non-C. elegans crowd, please give names of Drosophila homologs where relevant (e.g., when comparing to Drosophila expression patterns)

In a few places there are citations of popular science books or general textbooks (e.g., Carrol et al., 2004; Wolpert et al., 2019) . I think it would be better to cite review papers from the scientific literature or relevant primary papers.

We thank the four reviewers for their thoughtful and constructive comments. In light of their comments we made important revisions that are integrated in the final version, including also the correction of several typographical errors.

Article and author information

Author details

  1. Alison G Cole

    1. Department of Molecular Evolution and Development, University of Vienna, Vienna, Austria
    2. University of Vienna, Vienna, Austria
    Conceptualization, Data curation, Investigation
    Contributed equally with
    Tamar Hashimshony
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-7515-7489
  2. Tamar Hashimshony

    Department of Biology, Technion – Israel Institute of Technology, Haifa, Israel
    Conceptualization, Data curation, Visualization, Methodology
    Contributed equally with
    Alison G Cole
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-4786-838X
  3. Zhuo Du

    State Key Laboratory of Molecular Developmental Biology, Institute of Genetics and Developmental Biology, Chinese Academy of Sciences, Beijing, China
    Resources, Data curation
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-6322-4656
  4. Itai Yanai

    Institute for Computational Medicine, NYU School of Medicine, New York, United States
    Conceptualization, Formal analysis, Writing – original draft
    For correspondence
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-8438-2741


European Research Council (EvoDevoPaths)

  • Itai Yanai

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.


We thank the Technion Genome Center. This work was supported by a European Research Council grant (EvoDevoPaths). We are grateful for the EPIC2 dataset, which we used extensively. We also thank Yun Yan for work on an early version of this project.

Senior and Reviewing Editor

  1. Michael B Eisen, University of California, Berkeley, United States

Version history

  1. Preprint posted: February 5, 2023 (view preprint)
  2. Sent for peer review: February 21, 2023
  3. Preprint posted: May 5, 2023 (view preprint)
  4. Preprint posted: October 6, 2023 (view preprint)
  5. Version of Record published: January 29, 2024 (version 1)

Cite all versions

You can cite all versions using the DOI This DOI represents all versions, and will always resolve to the latest one.


© 2023, Cole, Hashimshony et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.


  • 340
    Page views
  • 51
  • 1

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Alison G Cole
  2. Tamar Hashimshony
  3. Zhuo Du
  4. Itai Yanai
Gene regulatory patterning codes in early cell fate specification of the C. elegans embryo
eLife 12:RP87099.

Share this article

Further reading

    1. Developmental Biology
    2. Evolutionary Biology
    Paul Knabl, Alexandra Schauer ... Grigory Genikhovich
    Research Article

    BMP signaling has a conserved function in patterning the dorsal-ventral body axis in Bilateria and the directive axis in anthozoan cnidarians. So far, cnidarian studies have focused on the role of different BMP signaling network components in regulating pSMAD1/5 gradient formation. Much less is known about the target genes downstream of BMP signaling. To address this, we generated a genome-wide list of direct pSMAD1/5 target genes in the anthozoan Nematostella vectensis, several of which were conserved in Drosophila and Xenopus. Our ChIP-seq analysis revealed that many of the regulatory molecules with documented bilaterally symmetric expression in Nematostella are directly controlled by BMP signaling. We identified several so far uncharacterized BMP-dependent transcription factors and signaling molecules, whose bilaterally symmetric expression may be indicative of their involvement in secondary axis patterning. One of these molecules is zswim4-6, which encodes a novel nuclear protein that can modulate the pSMAD1/5 gradient and potentially promote BMP-dependent gene repression.

    1. Biochemistry and Chemical Biology
    2. Developmental Biology
    Sima Stroganov ,  Talia   Harris  ... Michal Neeman
    Research Article

    Background: Fetal growth restriction (FGR) is a pregnancy complication in which a newborn fails to achieve its growth potential, increasing the risk of perinatal morbidity and mortality. Chronic maternal gestational hypoxia, as well as placental insufficiency are associated with increased FGR incidence; however, the molecular mechanisms underlying FGR remain unknown.

    Methods: Pregnant mice were subjected to acute or chronic hypoxia (12.5% O2) resulting in reduced fetal weight. Placenta oxygen transport was assessed by blood oxygenation level dependent (BOLD) contrast magnetic resonance imaging (MRI). The placentae were analyzed via immunohistochemistry and in situ hybridization. Human placentae were selected from FGR and matched controls and analyzed by immunohistochemistry (IHC). Maternal and cord sera were analyzed by mass spectrometry.

    Results: We show that murine acute and chronic gestational hypoxia recapitulates FGR phenotype and affects placental structure and morphology. Gestational hypoxia decreased labyrinth area, increased the incidence of red blood cells (RBCs) in the labyrinth while expanding the placental spiral arteries (SpA) diameter. Hypoxic placentae exhibited higher hemoglobin-oxygen affinity compared to the control. Placental abundance of Bisphosphoglycerate mutase (BPGM) was upregulated in the syncytiotrophoblast and spiral artery trophoblast cells (SpA TGCs) in the murine gestational hypoxia groups compared to the control. Hif1a levels were higher in the acute hypoxia group compared to the control. In contrast, human FGR placentae exhibited reduced BPGM levels in the syncytiotrophoblast layer compared to placentae from healthy uncomplicated pregnancies. Levels of 2,3 BPG, the product of BPGM, were lower in cord serum of human FGR placentae compared to control. Polar expression of BPGM, was found in both human and mouse placentae syncytiotrophoblast, with higher expression facing the maternal circulation. Moreover, in the murine SpA TGCs expression of BPGM was concentrated exclusively in the apical cell side, in direct proximity to the maternal circulation.

    Conclusions: This study suggests a possible involvement of placental BPGM in maternal-fetal oxygen transfer, and in the pathophysiology of FGR.

    Funding: This work was supported by the Weizmann Krenter Foundation and the Weizmann - Ichilov (Tel Aviv Sourasky Medical Center) Collaborative Grant in Biomedical Research, and by the Minerva Foundation (to MN), by the ISF KillCorona grant 3777/19 (to MN, MK).