Distinct gene expression dynamics in developing and regenerating crustacean limbs

Abstract

Regenerating animals have the ability to reproduce body parts that were originally made in the embryo and subsequently lost due to injury. Understanding whether regeneration mirrors development is an open question in most regenerative species. Here, we take a transcriptomics approach to examine whether leg regeneration shows similar temporal patterns of gene expression as leg development in the embryo, in the crustacean Parhyale hawaiensis. We find that leg development in the embryo shows stereotypic temporal patterns of gene expression. In contrast, the dynamics of gene expression during leg regeneration show a higher degree of variation related to the physiology of individual animals. A major driver of this variation is the molting cycle. We dissect the transcriptional signals of individual physiology and regeneration to obtain clearer temporal signals marking distinct phases of leg regeneration. Comparing the transcriptional dynamics of development and regeneration we find that, although the two processes use similar sets of genes, the temporal patterns in which these genes are deployed are different and cannot be systematically aligned.

Keywords: Parhyale hawaiensis; crustacean; leg development; regeneration; transcriptional profiling.

Fig. 1.

Transcriptional profiling of Parhyale leg development. (A) Morphology of Parhyale embryo and sampling of developing legs (E_f and E_d samples highlighted in gray and in blue, respectively). (B) Principal component analysis of the E_f and E_d series. PC1, representing 14% of the variance, correlates with developmental stage. (C) The developmental stage of the E_f and E_d samples is well predicted by RAPToR, using a reference built from the E_f samples (excluding the sample being tested; see Methods). (D) Heatmap representing the expression of 8,196 genes that are differentially expressed in the E_f and E_d time series. The dashed rectangle marks the developmental period that is covered by both the E_f and the E_d series.

Fig. 2.

Transcriptional profiling of Parhyale leg regeneration. (A) Morphology of Parhyale adult and sampling of regenerating legs (regenerating R_d and control R_p samples, highlighted in dark and light blue, respectively). The events of the different phases of regeneration, as established by live imaging (19), are indicated. (B–D) Principal component analysis of the R_d and R_p series. (B) PC1 separates the regenerating R_d samples from the pre-amputation (0 hpa), postmolt and control (R_p) samples. (C) Variation in PC2 is associated with the individual from which each sample was collected; R_d and R_p samples from the same individual show similar values (x axis, individuals ordered by time after amputation). (D) PC3 captures temporal changes that occur during regeneration in R_d, but not R_p. (E) PC1 of principal component analysis applied to the R_d series only, capturing temporal changes during regeneration. (F) Prediction of regenerative stage by RAPToR, using a reference built on the R_d series. To build the reference, fully differentiated legs (pre-amputation) were assigned to 300 hpa and the sample being tested was excluded (see Methods). RAPToR makes reasonable predictions of the stage of most R_d samples and matches most R_p samples with fully differentiated legs (300 hpa). The average distances between real time of collection and predicted time for the R_d and R_p samples are 30 and 170 h, respectively. Dev, deviance explained (gam regression, excluding the postmolt samples); r^2, r squared (linear correlation).

Fig. 3.

Impact of the molting cycle on the transcriptional profile of Parhyale legs. (A) Single T4 legs (dark gray) were sampled at different stages of the molting cycle: on the 5 d that precede molting (orange to yellow), 1 to 2 d postmolt (blue), and 9 to 10 d postmolt (purple). (B) Principal component analysis of these samples (large circles) captures molt-associated differences in PC1 and PC2. Projecting the R_d and R_p data on this PCA (in gray) reveals that the outliers of Fig. 2C (identified by number) were in the process of molting, whereas most other samples were in postmolt/intermolt phases (also see SI Appendix, Fig. S3.5). (C) Fuzzy c-means clustering of genes, based on expression values from the molting dataset, reported as centroid values. Three main transcriptional phases are observable, corresponding to postmolt/intermolt (clusters 6, 7, 1, 5, and 2), 5 to 3 d premolt (clusters 4 and 3), and 1 to 2 d premolt (cluster 8) periods.

Fig. 4.

Modeling of the regenerative signal. (A) Directed acyclic graph illustrating the model used to extract R values (dark red) from the raw counts of R_d and R_p series (gray circles). Gene levels in R_d samples (dark blue) are modeled as the product of gene levels in the corresponding R_p samples (light blue) multiplied by an R value (sampling error taken into account). (B) Principal component analysis of the R values: PC1 is strongly associated with the stage of regeneration. (C) Prediction of regenerative stage by RAPToR, using a reference built on the R series. To build the reference, fully differentiated legs (pre-amputation) were assigned to 300 hpa and the reference excluded the sample being tested (see Methods). Predictions are robust particularly in the early stages and they are largely independent of the gene set used to build the reference (SI Appendix, Fig. S5.2A). The average distance between real time of collection and predicted time for the R samples is 21 h. Dev, deviance explained (gam regression, excluding the postmolt samples); r^2, r squared (linear correlation).

Fig. 5.

Comparing the transcriptional dynamics of leg embryonic development and regeneration. (A) Combined principal component analysis of development (E series) and regeneration (R series); samples color-coded according to RAPToR pseudotimes. Variation in PC1 and PC2 is largely driven by embryonic development. (B) RAPToR temporal predictions on the R samples using a reference based on the E series. Coherent predictions are only made on pre-amputation and late or post-regeneration samples. Other stages are poorly predicted, and different sets of genes make incoherent predictions (SI Appendix, Fig. S5.2B). (C) Coexpression clusters defined by fuzzy c-means clustering of expression values in the developing (Left) and regenerating (Right) leg series. Four coexpressed gene clusters were identified in the E series (E2, E4, E1, and E3) and eight clusters were identified in the R series (R4, R1, R8, R2, R6, R3, R5, and R7). Heatmaps represent the average profiles (centroids) of each cluster. Clusters are ordered according to their temporal profiles (except clusters R3, R5, and R7, which do not show clear temporal profiles); samples are ordered by pseudotime. Cluster sizes are given in SI Appendix, Table S5. (D) Summary of the GO enrichment analysis for the E and R coexpression clusters; enriched GO terms were categorized as shown in SI Appendix, Fig. S5.7. (E) Number of genes shared between embryonic and regenerative coexpression clusters, expressed as a fold enrichment relative to equally sized random clusters. Clusters are ordered as in C and D (alternative ordering is shown in SI Appendix, Fig. S5.4). Similar results were obtained using clusters defined on untransformed E_f data (SI Appendix, Fig. S5.6). (F) Chord diagram depicting the genes shared between regenerative (Top) and embryonic (Bottom) coexpression clusters (aligned temporally from Left to Right). (Left) Diagram highlighting the genes of the R8 cluster (purple), corresponding to the regenerative phase of cell proliferation and patterning. (Right) Matches between all the regenerative and embryonic clusters. A fraction of genes (>5,000) are not clustered in the embryonic dataset. (G) Overlap of coexpressed gene clusters applied on a finer gene clustering of the E and R datasets as in E (see Methods). Alternative ordering of clusters is presented in SI Appendix, Fig. S5.9 and S5.10.

Fig. 6.

Temporal expression profiles of selected gene sets during leg development and regeneration. Expression in embryonic (E values, Left) and regenerating (R values, Right) legs, for genes associated with immune cells/responses (A), cell proliferation (B), patterning (C), differentiated nerves (D), and differentiated muscle (E). Samples ordered by pseudotime. t₀, pre-amputation; pm, postmolt.