Auxin-inducible degradation of mitotic regulators prevents binucleation

To investigate the function of binucleation, we developed a system in which we can specifically perturb intestinal endomitosis, without affecting other cell cycles within developing C. elegans larvae. We employed 2 mechanistically distinct approaches to block binucleation of the intestine; we either depleted CDK-1, which is essential for mitotic entry, or KNL-1, a conserved kinetochore protein that is required for chromosome segregation during M phase. CDK-1 inhibition is known to be a key mechanism to initiate endoreplication, and degrading CDK-1 essentially converts the endomitosis cycle to an endoreplication cycle, preventing mitotic entry and subsequent binucleation [14,17,20,21]. In contrast, KNL-1 inhibition does not affect mitotic entry but prevents chromosome segregation and partitioning of sister chromatids into 2 nuclei [22,23]. To deplete CDK-1 and KNL-1 specifically in intestinal cells and only during the time when endomitosis occurs, we made use of the auxin-inducible degradation system. In this system, proteins tagged with an auxin-inducible degron (AID) are degraded only in the presence of auxin and the F Box protein TIR1 [24–26]. By using a strain that expresses TIR1 under control of the intestine-specific Pges-1 promoter and exposing animals to auxin only during the time that endomitosis occurs, we can specifically inhibit intestinal binucleation without affecting canonical cell cycles of intestinal or other cells.

We generated AID knock-ins on the cdk-1 and knl-1 genes using CRISPR-mediated gene targeting and tested whether auxin induced depletion of CDK-1 or KNL-1 during endomitosis was able to block binucleation. In wild-type animals, endomitosis takes place at the end of the first larval stage in 12 to 14 of the 20 intestinal cells, resulting in 20 intestinal cells with 32 to 34 nuclei [19]. To control for nonspecific effects of auxin on development, we included an auxin control in each experiment, in which animals were treated with auxin for the same duration, but during a time in development in which CDK-1 or KNL-1 are not required in the intestine (Fig 1B and 1C). To assess the effect of auxin treatment on intestinal binucleation, we analyzed animals grown with auxin during endomitosis (hereafter referred to as auxin-treated animals or “+”) or in the auxin-control condition (hereafter referred to as auxin-control animals or “ct”) and found that auxin-treated animals contain 20 to 22 intestinal nuclei, in contrast to the 32 to 34 nuclei that are present in auxin-control animals (Fig 1D and 1E). Of the 12 to 14 intestinal cells that are known to undergo endomitosis and become binucleated in wild-type intestines, we found on average 13 cells to be binucleated in auxin-control animals compared to 0 binucleated cells in auxin-treated animals (Fig 1F and 1G). We next performed a quantitative DNA staining using PI and found that auxin-treated AID::cdk-1 and AID::knl-1 animals have a 64C DNA content, which is exactly double the nuclear ploidy of controls (Fig 1H). Thus, by inhibiting CDK-1 and KNL-1 during endomitosis, we can block binucleation and generate animals with mononucleated intestinal cells with the same cellular ploidy as wild-type binucleated intestinal cells.

Depletion of KNL-1 is known to prevent chromosome segregation in anaphase by blocking formation of kinetochore microtubule attachments [22]. To exclude that KNL-1 knockdown results in formation of micronuclei caused by chromosome missegregations, which can lead to DNA damage, cellular stress, and potentially arrest cells in the following cell cycle [27–30], we performed live-cell imaging of fluorescently labeled chromosomes in KNL-1 depleted cells. In all cells, KNL-1 depletion prevented chromosome segregation and resulted in mononucleation, but did not extend the duration of mitosis or result in the formation of micronuclei (S1A and S1B Fig, S1 and S2 Movies). After endomitosis, cells go through multiple rounds of endoreplication in which newly formed replicated chromosomes remain clustered together. Wild-type intestinal cells that have undergone endomitosis contain 2 nuclei, each containing 2 sets of chromosome clusters, which can be visualized by chromosomal LacO/LacI tagging (S1C Fig). As expected, we observed on average 4 chromosome clusters in nuclei of mononucleated KNL-1 depleted cells (S1D Fig), indicating that all copies of the labeled chromosome were present in a single nucleus. Finally, using a fluorescent cell cycle marker (Pges-1:: CYB-1^DB::mCherry), we found that preventing binucleation did not affect the timing of the subsequent endoreplicative S phase (S1E Fig). Taken together, our results indicate that KNL-1 depletion prevents binucleation without inducing detectable chromosomal or cell cycle aberrancies. Thus, this system provides a unique opportunity to study the function of binucleation, without altering the ploidy or number of cells in the tissue of interest.

Intestinal mononucleation decreases the nuclear surface-to-volume ratio, but does not affect cell size or morphology

To investigate whether mononucleation influences intestinal cell size or morphology, we used fluorescent markers in the AID::knl-1 strain to visualize the cell membranes and intestinal lumen. We found no effect of intestinal mononucleation on animal size, cell size, or lumen morphology (Fig 2A–2H). Because polyploidization has also been shown to influence nuclear morphology [31], we investigated the effect of binucleation on nuclear geometry using a fluorescently labeled nuclear pore protein (NPP-9::mCherry) as a marker for the nuclear membrane (Fig 2I and 2J). We measured nuclear size and found that the nuclear volume was increased with a factor of approximately 2.5 in mononucleated cells, indicating that per cell, the total nuclear volume has more than doubled (Fig 2K). The nuclear surface area was also increased in mononucleated cells, but to a lesser extent, with a factor of 2 (Fig 2L). Consequently, the surface-to-volume ratio decreased, indicating that less surface area is available per volume in mononucleated cells (Fig 2M). To test whether this effect is compensated to any extent by shape alterations or membrane invaginations that enlarge nuclear surface area, such has been observed in other polyploid cells [31], we also measured the ratio between the circumference and area of nuclear sections, but observed a similar effect (S2A–S2D Fig), indicating that there are no large invaginations that compensate for the amount of available nuclear surface area per volume. Thus, although mononucleation does not affect intestinal cell size or morphology, it produces nuclei that are more than twice as big and have an altered surface-to-volume ratio, which could potentially influence nuclear functions.

Download:

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

Fig 2. Intestinal cell mononucleation does not influence worm size, cell size, or intestinal morphology, but generates nuclei that are more than twice as big.

(A, B) Total body length (A) and width (B) of Pges-1::TIR1; AID::knl-1 young adults grown under control conditions (ct, n = 30) or in the presence of auxin (+, n = 32). Boxplots indicate the median and 25th to 75th percentile, error bars indicate min to max values, and individual values are shown as dots. NS, not significant (P > 0.05, Student t test). (C) Representative Z-stack projections of intestinal H2B-mCherry (DNA) and GFP-PH (membrane) in intestinal ring 3 of Pges-1::TIR1; AID::knl-1 animals grown on auxin or control conditions, used for quantifications of cell length and width. Scale bar is 20 μm. (D, E) Cellular length (D) and width (E) of Int3D/V cells of Pges-1::TIR1; AID::knl-1 young adults grown under control conditions (ct, n = 58) or in the presence of auxin (+, n = 24). Boxplots indicate the median and 25th to 75th percentile, error bars indicate min to max values, and individual values are shown as dots. NS, not significant (P > 0.05, Student t test). (F) Representative Z-stack projections of intestinal GFP-PH (membrane) in the apical membrane of intestinal ring 3 of Pges-1::TIR1; AID::knl-1 animals that were grown under control conditions or in the presence of auxin, used for quantifications of brush border and luminal width. Scale bar is 10 μm. (G, H) Intestinal brush border (G) and lumen (H) width in intestinal ring 3 of young adult worms grown under control (ct, n = 40) or auxin (+, n = 34) conditions. Boxplots indicate the median and 25th to 75th percentile, error bars indicate min to max values, and individual values are shown as dots. NS, not significant (P > 0.05, Mann–Whitney test). (I) Fluorescent images of intestinal NPP-9::mCherry localization in the nuclear membranes of Int3 cells of Pges-1::TIR1; AID::knl-1 adults grown under control or auxin conditions, corresponding to a 32C and 64C DNA content, respectively. Scale bar represents 10 μm. (J) Schematic depicting an ellipsoid shape and radii a, b, and c used to calculate nuclear volume and surface area. (K–M) Boxplots depicting the average nuclear volume (K), surface area (L) and surface-area-to-volume ratio (M) in binucleated (ct, n = 60 cells, 18 animals) or mononucleated (+, n = 56 cells, 30 animals) cells of Pges-1::TIR1; AID::knl-1 animals. Boxplots indicate the median and 25th to 75th percentile, error bars indicate min to max values, and individual values are shown as dots. P values were calculated by Mann–Whitney (K, L) or unpaired Student t test (M). Underlying data can be found in S1 Data. AID, auxin-inducible degron.

https://doi.org/10.1371/journal.pbio.3001597.g002

Binucleation is important for adult intestinal transcription

To investigate possible transcriptional differences between young adults with mononucleated or binucleated intestines, we performed single-worm RNA sequencing of AID::knl-1 and AID::cdk-1 auxin-treated and auxin-control animals. Since the intestine is one of the most transcriptionally active and largest tissues in the worm, making up roughly one-third of the animals volume [32], we anticipated that changes in transcription in the intestine would be detectable in whole worm sequencing. We performed differential gene expression analysis for 68 (AID::knl-1) and 75 (AID::cdk-1) young adult (72 hours) auxin-treated or auxin-control animals (see Methods for details). For the AID::knl-1 animals alone, 16 percent of genes included in the analysis showed significant and substantial differential expression between worms with a mononucleated or binucleated intestine (420/2,638 genes, Fig 3A, S3A Fig). Sequencing of AID::cdk-1 animals revealed fewer differentially expressed genes between animals with mononucleated or binucleated intestines, which is likely due to a lower complexity of these samples (S3B–S3E Fig). Although it is unclear why the AID::cdk-1 samples had lower complexities, we focused on the overlap in differential gene expression between the AID::knl-1 and AID::cdk-1 strains to identify genes that are differentially expressed in animals with mononucleated intestines (S3F Fig). We found a strong enrichment (P = 2e-7) in the overlap between genes significantly and substantially down-regulated in both strains, consisting of 15 genes (Fig 3B and 3C). One of the genes that stood out was vit-2, 1 of the 6 C. elegans vitellogenin genes whose levels have been shown to correlate strongly with the growth and fitness traits of C. elegans progeny [33]. Vitellogenins are highly expressed in adult intestines and are essential to mobilize lipids in the intestine for transport to the developing oocytes in the germline, where they contribute to progeny development [34–37]. Because vitellogenins have previously been shown to function redundantly, and down-regulation of individual vit genes often leads to up-regulation of others [34], we analyzed the expression of all 6 vit genes in our dataset and found that all of them were down-regulated in worms with a mononucleated intestine (S4A–S4F Fig). Moreover, the sum of all vit gene expression values showed a stronger decrease than any gene alone (S4G Fig), suggesting that in mononucleated animals, the down-regulation of vit-2 is not compensated by the transcriptional up-regulation of other vitellogenins.

Download:

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

Fig 3. Blocking binucleation causes transcriptional down-regulation of intestinally expressed genes in young adult C. elegans.

(A) Volcano plot of RNA sequencing data depicting the transcriptional gene up- and down-regulation in worms with a mononucleated intestine, compared to worms with a binucleated (wild type) intestine in Pges-1::TIR1; AID::knl-1 animals (n = 68). After filtering for coverage, batch consistency, and intestinal expression, 2,638 genes were analyzed for differential gene expression using DEseq. Red dots represent genes significantly differentially expressed (adjusted P value < 0.05, top 25% absolute log2(foldchange)). Genes significantly differentially expressed in both Pges-1::TIR1; AID::knl-1 and Pges-1::TIR1; AID::cdk-1 comparisons were individually annotated with their gene name (excluding genes without a gene name). (B) Venn diagram of the overlap of significantly down-regulated genes in Pges-1::TIR1; AID::knl-1 and Pges-1::TIR1; AID::cdk-1 animals. (C) Overview of genes that are found to be significantly down-regulated in worms in which intestinal binucleation is blocked, both by degradation of CDK-1 and degradation of KNL-1, including a short description and previously established associated phenotypes as described on WormBase [38]. Underlying data are available at the Gene Expression Omnibus, identifier GSE169330, and in S1 Data. AID, auxin-inducible degron.

https://doi.org/10.1371/journal.pbio.3001597.g003

Binucleation of the intestine promotes vitellogenin expression and lipid loading into oocytes

To confirm that vit-2 expression is down-regulated in animals with mononucleated intestinal cells, we generated a Pvit-2::^NLSGFP transcriptional reporter to measure vit-2 promoter activity at different moments of development. Since vitellogenin expression is absent during larval development and drastically up-regulated at the L4-to-adult transition [34], expression levels are still relatively low at 48 hours of development, around the end of the L4 stage (Fig 4A). Upon adulthood, vit-2 expression levels increase considerably in both auxin-treated and auxin-control animals. However, vit-2 promoter activity shows a significant reduction in auxin-treated animals at 72 hours of development. This difference in vit-2 expression levels is no longer present in older adults, when vit-2 expression is peaking and becomes similar between animals with mononucleated and binucleated intestinal cells (Fig 4A, 96 and 120 hour time points). To further investigate whether mononucleated intestinal cells have reduced ability to express vit-2 during early stages of adulthood, we performed single molecule fluorescence in situ hybridization (smFISH) at 54 hours of development, when vit-2 starts becoming expressed in intestinal cells and it is possible to count single vit-2 mRNA molecules in intestinal cells. Our analyses revealed that animals with mononucleated intestinal cells had on average fewer vit-2 mRNAs per cell and reduced nascent transcription in their nuclei (Fig 4B–4D). Consistent with a transcriptional down-regulation of vit-2 in animals with mononucleated intestinal cells, we also found lower levels of VIT-2 protein in embryos derived from mothers with mononucleated intestines (Fig 4E). Again, VIT-2 levels were significantly lower in embryos derived from auxin-treated 72-hour old adults and, similar to the vit-2 promoter activity, the difference between auxin-treated and control animals was no longer present on subsequent days of adult development.

Download:

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

Fig 4. Perturbation of binucleation decreases yolk protein expression and lipid transport to the germline.

(A) Fluorescence images and normalized total fluorescence intensities of a Pvit-2::^NLSGFP transcriptional reporter at different moments during adult development for Pges-1::TIR1; AID::knl-1 animals grown under control (ct, n = 53 to 111) or auxin (+, n = 71 to 111) conditions, in 5 replicate experiments. Each data point represents the normalized fluorescence intensity of one worm. Boxplots indicate the median and 25th to 75th percentile, error bars indicate min to max values, and individual values are shown as dots. P values were calculated by Student t test. NS, not significant (P > 0.05). (B–D) smFISH analyses of vit-2 mRNAs in Pges-1::TIR1; AID::knl-1 animals at 54 hours of development, grown under control (ct) or auxin (+) conditions. (B) Z-stack projections of vit-2 smFISH (magenta) in Int3D/V cells. Animals contain an intestinal membrane marker (Pges-1::GFP-PH) that was used to determine cell boundaries (dashed line). Scale bar represents 10 μm. (C) Quantification of number of cellular vit-2 mRNAs in Int3D/V cells of animals grown under control (ct, n = 62) or auxin (+, n = 88) conditions in 2 replicate experiments. Scatter plot with median (middle line) and 95% confidence interval (error bars). Each dot represents the number of vit-2 mRNAs in a single cell. P value was calculated by Mann–Whitney test. (D) Quantification of number of nascent vit-2 transcripts per cell in Pges-1::TIR1; AID::knl-1 animals grown under control (ct) or auxin (+) conditions. Scatter plot with median (middle line) and 95% confidence interval (error bars). Each dot represents the number of nascent vit-2 mRNAs in a single cell. P value was calculated by Mann–Whitney test. (E) Fluorescence images and boxplots showing total endogenous VIT-2::GFP fluorescence intensity in early embryos (1-cell stage to 4-cell stage) derived from control (ct, n = 38 to 71) or auxin (+, n = 36 to 71) Pges-1::TIR1; AID::knl-1 animals, in 2 replicate experiments. Boxplots indicate the median and 25th to 75th percentile, error bars indicate min to max values, and individual values are shown as dots. P values were calculated by Mann–Whitney test. NS, not significant (P > 0.05). (F) Z-stack projection of fluorescence images and boxplots of normalized total fluorescence intensities of BODIPY lipid staining of early embryos isolated from young adult (72 hours) Pges-1::TIR1; AID::knl-1 animals grown under control (ct, n = 56) or auxin (+, n = 77) conditions in 3 replicate experiments. Boxplots indicate the median and 25th to 75th percentile, error bars indicate min to max values, and individual values are shown as dots. P values were calculated by Mann–Whitney test. (G) Z-stack projection images and boxplots of normalized fluorescence intensities of BODIPY lipid staining of Pges-1::TIR1; AID::knl-1 young adults (72 hours) that were grown under control (ct, n = 62) or auxin (+, n = 63) conditions in 2 replicate experiments. Boxplots indicate the median and 25th to 75th percentile, error bars indicate min to max values, and individual values are shown as dots. P values were calculated by Mann–Whitney test. Underlying data can be found in S1 Data. AID, auxin-inducible degron; smFISH, single molecule fluorescence in situ hybridisation.

https://doi.org/10.1371/journal.pbio.3001597.g004

Because vitellogenins are required to transport lipids from the intestine to the germline, we investigated whether decreased vitellogenin levels resulted in a reduction of lipid loading into oocytes. For this, we used BODIPY staining to quantify lipid levels in embryos and young adult worms. We observed lower lipid levels in embryos from auxin-treated worms, consistent with reduced lipid loading into oocytes of animals with mononucleated intestines (Fig 4F). Moreover, we found higher amounts of lipids in the intestines of adults with mononucleated intestines (Fig 4G), indicating that specifically the transport of lipids is impaired, rather than lipid production or uptake. The effect of mononucleation on lipid levels was more striking than the decrease that we observed in Pvit-2::^NLSGFP expression, which is consistent with the notion that vitellogenins function in 2 distinct yolk complexes, and that multiple vitellogenins are down-regulated in mononucleated animals. Together, our findings indicate that expression of vitellogenins is hampered by blocking binucleation in the intestine, resulting in reduced lipid loading into developing oocytes.

Intestinal binucleation enhances C. elegans fitness

Abundant expression of vitellogenins in the adult maternal C. elegans intestine is important to support postembryonic development and fertility of the offspring [33,36]. To assess whether intestinal binucleation is functionally important to generate sufficient vitellogenins to promote progeny fitness, we performed a competitive fitness assay with animals that have either mononucleated or binucleated intestinal cells. To this end, we generated strains with either a Pmyo-2::mCherry or a Pmyo-2::GFP pharyngeal marker in addition to the AID::knl-1 or AID::cdk-1 alleles. This allowed us to follow the progeny of animals with mononucleated or binucleated cells over several generations. By mixing equal amounts of animals with mononucleated (auxin-treated, +) or binucleated intestines (control condition, ct) on plates and counting the proportion of GFP and mCherry positive progeny after several generations, we could determine how binucleation influences reproductive fitness (Fig 5A). It is important to note that auxin was only added during development of the parental worms and that progeny were not treated with auxin and their intestinal cells were thus binucleated. To control for fitness differences due to the Pmyo-2::mCherry or Pmyo-2::GFP markers, we included experiments with reversed fluorescent markers and used measurements of the fitness difference between Pmyo-2::mCherry and Pmyo-2::GFP control worms to normalize our data (see Methods for details). Using the AID::knl-1 strain, we found that after one generation, animals with a wild-type intestine had given rise to an average of 57.0% (±14.4%) of the population, while animals with a mononucleated intestine only gave rise to 43.0% (±14.4%) of the population (Fig 5B). Similar numbers were obtained with the AID::cdk-1 strain (Fig 5C). These results demonstrate that animals with binucleated intestines have a significant fitness advantage over animals with mononucleated intestines.

Download:

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

Fig 5. Inhibition of intestinal binucleation reduces relative reproductive fitness.

(A) Overview of the relative fitness assay. L4 animals grown under auxin (+) or control (ct) conditions carrying a pharyngeal marker in either red (Pmyo-2::mCherry) or green (Pmyo-2::GFP) are transferred in equal amounts to a single plate. Worms are grown until the plate is full or starved, after which a random subset of worms are transferred to a new plate, effectively diluting the population, after which the proportion of mCherry+ and GFP+ progeny is counted. In replicate experiments, pharyngeal markers were swapped for the auxin and control conditions. Reproductive fitness measurements shown in panels B and C were made after one plate dilution, whereas reproductive fitness measured over consecutive plate dilutions is shown in panel G. (B, C) Boxplots showing the normalized proportion of progeny originating from worms with a wild-type binucleated (ct, n = 40 plates/280 worms and 34 plates/238 worms) or mononucleated (+, n = 40 plates/280 worms and 34 plates/238 worms) intestine in relative fitness assays using either the AID::cdk-1 (B) or AID::knl-1 alleles (C) to block binucleation, in 2 replicate experiments. Measurements were made after one plate dilution. Boxplots indicate the median and 25th to 75th percentile, error bars indicate min to max values, and individual values are shown as dots. (D) Boxplot showing total brood size of Pges-1::TIR1; AID::knl-1 animals grown under control conditions (ct, n = 43 plates) or in the presence of auxin (+, n = 43 plates), in 3 replicate experiments. Amounts of progeny were counted for 3 days of egg laying. Boxplots indicate the median and 25th to 75th percentile, error bars indicate min to max values, and individual values are shown as dots. (E) Violin boxplots depicting progeny growth rates from worms with increasing maternal age, grown under control conditions (ct, n = 9 plates) or in the presence of auxin (+, n = 9 plates), in 3 replicate experiments. For each condition, the timing of the L3 molt was quantified (see Methods for details), and the average timing in the control condition for each maternal age was used to calculate the developmental delay. Horizontal lines indicate the median and 25th to 75th percentile, violin plots extend to min and max values and individual values are shown as dots. (F) Scatter plots of brood size in the second generation of Pges-1::TIR1; AID::knl-1 animals (F2) that originated from animals grown under control (ct, n = 70 plates) or auxin (+, n = 71 plates) conditions. Amounts of animals were counted for 3 days of egg laying. Line and error bars represent mean and SEM. (G) Exponential growth model assuming a intergenerational (green) or transgenerational (orange) effect of blocking binucleation on reproductive fitness. Measurements of the mean reproductive fitness that were obtained over consecutive plate dilutions are shown in black, and error bars depict SEM. P values were calculated by Mann–Whitney test. NS, not significant (P > 0.05). Underlying data can be found in S1 Data. AID, auxin-inducible degron.

https://doi.org/10.1371/journal.pbio.3001597.g005

To understand how perturbation of binucleation affects fitness, we investigated several aspects of worm growth and reproduction upon inhibition of binucleation. When analyzing AID::knl-1 and AID::cdk-1 animals, we noticed that the AID::cdk-1 strain had smaller brood sizes and increased embryonic lethality compared to wild-type animals, both in the presence or absence of auxin, suggesting that the AID tag compromises CDK-1 function resulting in a weak hypomorph (S5A–S5C Fig). Nonetheless, when comparing animals with mononucleated or binucleated intestines in either the AID::cdk-1 or AID::knl-1 background, we found no reduction in brood sizes in animals with mononucleated intestines (Fig 5D, S5C Fig). However, we did find a significant delay in progeny growth when examining eggs that were laid between 72 and 96 hours of development, corresponding to the first and second day of adulthood (Fig 5E). This delay was not due to growing animals on auxin or the presence of TIR1, as worms lacking the AID::knl-1 or AID::cdk-1 alleles did not show developmental delays or reproductive defects (S5D and S5E Fig). Moreover, animals derived from mothers with mononucleated intestines showed a mild decrease in brood sizes compared to controls (Fig 5F). To test whether these developmental effects could account for the observed decreases in fitness, we used an exponential growth model to predict the relative fitness of worms with a mononucleated intestine based on our growth and reproduction measurements. In this model, we differentiated the possibilities of an intergenerational effect of blocking binucleation in the intestine, where only the first generation of progeny is affected, and a transgenerational effect that lasts for several generations. Comparing this model with our experimental data revealed 2 things. First, the plateau in the proportion of progeny coming from mothers with a mononucleated intestine indicates that there is an intergenerational rather than a transgenerational effect on progeny fitness (Fig 5G). Second, the fitness decrease that we measured closely matches the model, suggesting that decreases in progeny growth and reproduction fully explain the difference in relative fitness. Taken together, our data show that intestinal binucleation is important for reproductive fitness and that blocking binucleation decreases progeny growth and reproduction.

Binucleation of the C. elegans intestine is important for the rapid up-regulation of gene expression

To understand why mononucleated cells have decreased levels of vit-2 gene expression compared to binucleated cells, we performed in-depth analysis of our single-worm RNA sequencing data to identify similarities between genes that are affected by binucleation. First, we found no correlation between expression levels and differential gene expression in animals with a mononucleated intestine, indicating that binucleation is not solely important for the expression of highly or lowly expressed genes (S3A and S3C Fig). Next, we used sequencing data from wild-type worms at different stages of development, available from ModEncode [38], and analyzed the developmental expression profiles of the genes that we identified in our sequencing of animals with mononucleated intestinal nuclei. Overall, genes that are down-regulated in worms with a mononucleated intestine show an increase in gene expression throughout wild-type development, with a strong increase in expression from the L4 to adult stage (Fig 6A). In contrast, genes either unaffected by perturbation of binucleation or genes that are intestinally expressed showed a more constant expression profile during wild-type development. In addition, we found a substantial enrichment (P = 6.4e-8) of down-regulated genes located on the X chromosome (S6A Fig). The X chromosome–enriched down-regulation is not restricted to vitellogenin genes, of which 5 out of 6 are located on the X chromosome, but rather a global repression of all X chromosomal genes (S6B Fig). Possibly, an altered X chromosome organization within the nucleus contributes to the decreased expression in mononucleated cells.

Download:

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

Fig 6. Binucleation allows rapid up-regulation of transcription.

(A) Mean expression for down-regulated (n = 96 genes), unaffected (n = 113 genes), or intestinally expressed (n = 80 genes) subsets of genes in different developmental stages from ModEncode FPKM gene expression data. Area around curve shows confidence interval. (B) Schematic overview of heat shock–inducible NLS-GFP transgene. smFISH probes were designed against the sfGFP sequence. (C) Schematic overview of smFISH and imaging experiments following heat shock induction in Pges-1::TIR1; AID::knl-1 animals. Young L4 stage animals grown under control or auxin conditions were heat shocked in a water bath at 33°C for 30 minutes, followed by 5 minutes in a 37°C air incubator. Animals were left to recover at 25°C. Samples were taken after 15 minutes for smFISH analyses or every hour, for the analysis of nuclear GFP accumulation. (D) Total nuclear fluorescence intensities of intestinal nuclei at different time points after heat shock for animals grown under control (ct, n = 13 to 44) or auxin (+, n = 18 to 35) conditions, in 3 replicate experiments. Boxplots indicate the median and 25th to 75th percentile, error bars indicate min to max values, and individual values are shown as dots. P values were calculated by Student t test. (E) Representative smFISH images and quantifications of cellular GFP mRNA concentration in animals grown under auxin (+, n = 113) or control (ct, n = 112) conditions 15 minutes after heat shock, in 2 replicate experiments. Scatter plot with median (middle line) and 95% confidence interval (error bars). Each dot represents mRNA concentration in a single cell. P value was calculated by Mann–Whitney test. Scale bar represents 5 μm. Underlying data can be found in S1 Data. AID, auxin-inducible degron; FPKM, fragments per kilobase of exon per million mapped fragments; smFISH, single molecule fluorescence in situ hybridisation.

https://doi.org/10.1371/journal.pbio.3001597.g006

To test whether binucleation is important for the rapid up-regulation of gene expression in general, or specifically for expression of genes that are up-regulated upon adulthood, we measured the up-regulation of a heat shock–inducible GFP-NLS reporter in animals with mononucleated or binucleated intestines. In these experiments, L4 stage animals were heat shocked, and nuclear GFP intensities were measured at multiple time points after heat shock (Fig 6B and 6C). Interestingly, mononucleated intestinal cells showed a delay in the up-regulation of nuclear GFP after heat shock compared to binucleated controls, whereas the accumulation of GFP signal in body wall muscle nuclei was similar between the 2 conditions (Fig 6D, S7 Fig). To confirm that this delay in up-regulation of heat shock–inducible expression arises at the transcriptional level, we quantified the cellular GFP mRNA density 15 minutes after heat shock induction using smFISH. Consistent with a delay in GFP up-regulation, we observed a significant decrease in GFP mRNA levels in mononucleated compared to binucleated intestinal cells (Fig 6E). Together, these results show that rapid up-regulation of transcription is impaired in cells with one rather than two nuclei.

Increasing vitellogenin levels can rescue the defects in the progeny of animals with mononucleated intestinal cells

Our results suggest that animals with mononucleated intestinal cells have a defect in the rapid up-regulation of vitellogenin gene expression at the L4-to-adult transition, leading to reduced lipid loading in their progeny and reduced fitness. To test whether reduced vitellogenin expression can phenocopy the loss of endomitosis, we generated knockout mutants in vit-5 and vit-6, 2 of the 6 vitellogenin genes. Single and double vit-5 and vit-6 mutants had lower brood sizes, and brood sizes were further reduced in the progeny of animals with mononucleated intestines (Fig 7A). This is consistent with previous studies, where it was found that depletion of the yolk uptake receptor RME-2 results in reduced fertility [33,35]. Thus, reduction of yolk proteins gives rise to similar defects as the loss of endomitosis.

Download:

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

Fig 7. Transient activation of vitellogenin expression rescues defects of animals with mononucleated intestines.

(A) F2 brood sizes of Pges-1::TIR1; AID::knl-1 animals containing deletions in either vit-5, vit-6, or both. F2 brood sizes were measured from the progeny of animals with either mononucleated (+, n = 43 to 44) or binucleated (ct, n = 41 to 44) intestines that had been laid during the first day of adulthood (maternal age 48–72), in 2 replicate experiments. Animals without vit-5 or vit-6 deletions (wild type) were taken along as controls. (B) Normalized total Pvit-2::^NLSGFP fluorescence intensity per animal, for animals grown under auxin (+) or control (ct) conditions and with (+, n = 17 and 29) or without (−, n = 82 and 89) intestinal overexpression of transcription factors ceh-60 and unc-62, measured at young adulthood in 5 replicate experiments. Scatter plot with line and error bars indicating mean and SEM. (C) Normalized total fluorescence intensities of BODIPY lipid staining of early embryos isolated from young adult (72 hours) Pges-1::TIR1; AID::knl-1 animals with or without intestinal overexpression of CEH-60 and UNC-62, grown under control (ct, n = 76 and 64) or auxin (+, n = 67 and 64) conditions in 3 replicate experiments. Boxplots indicate the median and 25th to 75th percentile, error bars indicate min to max values, and individual values are shown as dots. (D) F2 brood sizes of Pges-1::TIR1; AID::knl-1 animals with or without intestinal overexpression of CEH-60 and UNC-62. F2 brood sizes were measured from the progeny of animals with either mononucleated (+, n = 42 and 42) or binucleated (ct, n = 40 and 41) intestines that had been laid during the first day of adulthood, in 2 replicate experiments. P values were calculated by Mann–Whitney test. NS, not significant (P > 0.05). Underlying data can be found in S1 Data. AID, auxin-inducible degron.

https://doi.org/10.1371/journal.pbio.3001597.g007

We next tested whether transient activation of vitellogenin transcription earlier in development is sufficient to mitigate the differences in vit-2 expression between binucleated and mononucleated intestines and can rescue the defects that we observe in the progeny of animals with mononucleated intestinal cells. During wild-type development, vitellogenin expression is up-regulated at the L4-to-adult transition by the interaction between transcription factors CEH-60 and UNC-62 [39,40]. We found that intestine-specific overexpression of CEH-60 and UNC-62 was sufficient to activate expression of our vitellogenin reporter as early as the L3 stage, whereas no activation was observed at this stage in control animals (S8 Fig). Furthermore, animals overexpressing CEH-60 and UNC-62 showed an increase in Pvit-2 expression and their progeny had inherited increased levels of lipids (Fig 7B and 7C), consistent with the function of CEH-60 and UNC-62 to drive vitellogenin transcription and promote lipid loading into oocytes. Importantly, overexpression of CEH-60 and UNC-62 resulted in similar Pvit-2 expression levels between mononucleated and binucleated intestines, as well as similar levels of lipids in embryos derived from mothers with mononucleated or binucleated intestines (Fig 7B and 7C). These results suggest that CEH-60 and UNC-62 are limiting for vitellogenin expression in mononucleated intestinal cells and that overexpression of these factors is sufficient to rescue the defects in lipid loading in the progeny of animals with mononucleated intestinal cells.

Finally, we tested whether overexpression of CEH-60 and UNC-62 was sufficient to rescue the reduced brood sizes of progeny derived from animals with mononucleated intestines. Although overexpression of CEH-60 and UNC-62 lead to smaller broods compared to wild type, we no longer found differences in the brood sizes of the progeny of animals with either binucleated or mononucleated intestinal cells (Fig 7D). Thus, increasing vitellogenin levels by overexpression of CEH-60 and UNC-62 is sufficient to abrogate the phenotypic consequences of reduced transcription in animals with mononucleated intestines. Taken together, these results indicate that binucleation of the intestine is needed to facilitate rapid transcriptional up-regulation of vitellogenin expression at the L4-to-adult transition, which is important to ensure abundant lipid loading into developing oocytes and promote reproductive fitness.

Worm culture

Worms were cultured on nematode growth medium (NGM) plates seeded with OP50 Escherichia coli bacteria according to standard protocols. All strains were maintained at 20°C, unless indicated otherwise. N2 Bristol were used as wild types. For auxin plates, NGM was supplemented with 0.5 mM auxin (I2886; Sigma-Aldrich, St. Louis, MO, USA) after autoclaving and cooling to <55°C. As auxin is dissolved in 100% ethanol, control plates are supplemented with equal volumes of 100% ethanol without auxin. For auxin experiments, worms were synchronized by isolation of eggs from adult hermaphrodites through alkaline hypochlorite treatment and hatching overnight in M9 medium in the absence of food. Next, synchronized L1 larvae were divided over 2 conditions: auxin (+) and auxin control (ct). For the auxin condition, worms were grown on auxin for 24 hours, after which they are transferred to control plates. Auxin-control worms were grown on control plates for 24 hours and transferred to auxin plates from 24 to 48 hours of development. For fitness assays, worms were grown on NGM plates supplemented with 100 mg/mL ampicillin (Sigma-Aldrich A-9518) L4440 E. coli bacteria to prevent contamination. For heat shock experiments, animals were heat shocked in a water bath at 33°C for 30 minutes, followed by 5 minutes in a 37°C air incubator. Animals were then transferred to 25°C, and samples were taken after 15 minutes for smFISH analysis (see below) or every hour, for the analysis of nuclear GFP accumulation.

Strains

C. elegans strains used in this study are described in S1 Table. CDK-1 and KNL-1 N-terminal AID tags were inserted into the CA1209 strain using CRISPR/Cas-9 genome editing with 2 overlapping ssODNs (IDT 4 nmole Ultramers) as template for homology-directed repair, as described by Paix and colleagues [55]. vit-5(mat169) and vit-6(mat171) deletions were generated using CRISPR/Cas-9 genome editing by using two 3′ guides and two 5′ guides and a rol-6 co-injection marker. CRISPR/Cas-9 insertions and deletions were selected by genomic PCRs and confirmed by DNA sequencing. Single guide RNA and ssODN repair template sequences are listed in S2 Table. Intestinal markers (matIs53), a nuclear membrane marker (matIs104), pharyngeal markers (matIs114 and matIs116), and an S/G2-phase marker were integrated using gamma irradiation and outcrossed 4 times. The transcriptional Pvit-2::sfGFP reporter strain (matIs137) was integrated using CRISPR/Cas-9 mediated genome editing at the MosI insertion allele ttTi5605 on chromosome II. For the intestinal CEH-60 and UNC-62 overexpression experiments, 2 strains were generated: an extrachromosomal overexpression strain, containing Pelt-2::ceh-60; Pges-1::BFP-P2A-unc-62; Pmyo-2::mCherry (matEx151, used in Fig 7B), and an integrated overexpression strain, containing Pelt-2::ceh-60; Pges-1::BFP-P2A-unc-62; Plin-48::TdTomato (matIs172, used for experiments in Fig 7C).

Microscopy and image analysis

Static and time-lapse epifluorescence imaging of intestinal fluorescent markers, nuclear markers, chromosome markers, vitellogenin levels, and DNA/lipid staining was performed on a Ultraview Vox spinning disk confocal microscope (Perkin Elmer, Waltham, MA, USA) equipped with a CMOS digital camera (C13440; Hamamatsu Photonics, Hamamatsu-city, Japan). Differential interference contrast (DIC) imaging of worm size was performed using a 20× objective on a Axio Imager M2 microscope (Carl Zeiss, Oberkochen, Germany) equipped with an Axiocam camera (Carl Zeiss, Oberkochen, Germany). Imaging of cell cycle progression using the S/G2-phase marker was performed on a DM6000 Fluorescence Microscope (Leica Microsystems GmbH, Wetzlar, Germany). For imaging, animals were mounted onto 2% agarose pads (adults) or 7% agarose pads (L1 and L2 larvae) and immobilized with 10 mM sodium azide for static imaging, or 1 ng/mL levamisole and 100 ng/mL tricaine for time-lapse imaging. Young adult intestinal markers or nuclear membrane markers were imaged with a 20× or 63× objective, using 2× binning. Chromosome markers in L2/L3 animals were imaged with a 100× objective, without binning. Time-lapse imaging of intestinal markers during endomitosis in L1 larvae was performed with a 100× objective, with 2× binning. For imaging in the presence of auxin, M9 and agarose pads were supplemented with 0.5 mM auxin. To measure vitellogenin and lipid levels, embryos were imaged at 100× magnification, using 2× binning. PI DNA staining was imaged with a 63× objective, with 2× binning.

Images were processed and analyzed using Fiji software [56]. For quantifications of intestinal cell and nuclei numbers, images of young adults (64 hours) were taken using a 20× objective lens and stitched together using the pairwise stitching tool from Fiji [57]. For quantifications of PI stainings, GFP-VIT-2 fluorescence intensity, and BODIPY stainings, z-axis serial scans distanced at 50% the optical slice depth were summed in a z-stack projection after background subtraction, after which the integrated density of the region of interest was measured. For quantifications of nuclear GFP after heat shock induction, 20 z-slices with 0.5 μm spacing (corresponding to 50% of the optical slice depth) were summed in a z-stack projection. In this case, no background subtraction was performed, as background intensity analysis showed similar background levels between conditions. From the z-projections, integrated densities of the nuclear regions were measured. For binucleated cells, the integrated densities of the 2 nuclei were summed.

DNA staining

For PI quantitative DNA staining, freeze-cracked young adult hermaphrodites were fixed with Carnoy’s solution and treated with RNAse A (Invitrogen #12091021; Thermo Fisher Scientific, Waltham, MA, USA) as previously described [58]. PI staining was performed using 100 μg/mL PI (#P4170; Sigma-Aldrich, St. Louis, MO, USA) for 90 minutes at 37°C, after which slides were washed 3 times and mounted using ProLong gold antifade mountant (Invitrogen #P36934).

Competitive fitness assay and modeling

Competitive fitness assays were performed using the AID::cdk-1 strains GAL182 (with Pmyo-2::GFP) and GAL191 (with Pmyo-2::mCherry) and the AID::knl-1 strains GAL160 (with Pmyo-2::mCherry) and GAL162 (with Pmyo-2::GFP). Prior to the start of the assay, worms were grown under auxin or auxin-control conditions as described above. Each experiment contained 3 conditions: control-GFP versus control-mCherry, auxin-GFP versus control-mCherry, and control-GFP versus auxin-mCherry. In the first condition, control-GFP versus control-mCherry, auxin-control worms with a green fluorescent pharyngeal marker were grown on the same plate as auxin-control worms with a red fluorescent pharyngeal marker. Second, for the auxin-GFP versus control-mCherry condition, auxin worms with a green fluorescent pharyngeal marker were grown on the same plate as auxin-control worms with a red fluorescent pharyngeal marker. Finally, in the control-GFP versus auxin-mCherry condition, auxin-control worms with a green fluorescent pharyngeal marker were grown on the same plate as auxin worms with a red fluorescent pharyngeal marker. For each of the conditions, 10 replicates were performed. In each replicate, 7 worms of both conditions were transferred to single plates, which were left to grow until starvation (for AID::knl-1 worms) or until the plate was close to starvation (for AID::cdk-1 strains, as these were prone to become sterile upon starvation). Each plate was chunked to 2 new plates, one of which was grown for 1 day and used to count relative amounts of GFP+ and mCherry+ progeny, whereas the other plate was used to initiate another growth cycle. For counting, animals were washed off plates, mounted on 2% agarose pads containing 10 mM sodium azide, and counted using a Zeiss Axioscope. Relative amounts of control-GFP versus control-mCherry progeny were used to normalize relative amounts of auxin-GFP versus control-mCherry and control-GFP versus auxin-mCherry progeny.

For modeling, the average number of eggs and average time of egg laying were calculated for the first and second generation, using a predicted Gaussian distribution curve on measurements of eggs laid per day. These numbers were used to calculate an exponential growth rate λ, equal to , where x is the average number of eggs that one animal produces, and t is the average time (in days) at which these eggs are laid. The predicted relative fitness change per day of growth was determined by the ratio of the growth rate of worms with a mononucleated intestine divided by the growth rate of worms with a binucleated intestine. In the model assuming a transgenerational effect, this change in relative fitness is constant for several generations. When assuming a intergenerational effect, the change in relative fitness is set to zero from the start of generation F2.

Single-worm RNA sequencing and analysis

Transgenic animals were grown until adulthood under auxin or control conditions. Young adult hermaphrodites were picked using tweezers and washed in 0.5-mL sterile demineralized water before adding 200-μL trizol (Invitrogen 15596018). Samples were stored at −80°C for a limited time. mRNA extraction, barcoding, reverse transcription, in vitro transcription, and Illumina sequencing library preparation were performed according to the robotized version of the CEL-seq2 protocol [59] using the SuperScript II Double-Stranded cDNA synthesis kit (Thermo Fisher Scientific, Waltham, MA, USA), Agencourt AMPure XP beads (Beckman Coulter, Brea, CA, USA), and randomhexRt for converting aRNA to cDNA using random priming. The libraries were sequenced paired-end at 50 bp read length on an Illumina HiSeq 2500. The 50 base pair paired-end reads were aligned to the C. elegans reference transcriptome compiled from the C. elegans reference genome WS249. Raw data were processed using R (v3.6.2). Samples were filtered to include those with sufficient counts per transcript and gene, results were filtered for batch consistency, and only genes that were found at least once in every sample were included for downstream analysis (S3 Fig). Further filtering of genes was performed to include only genes for which intestinal expression has been annotated in previous analyses of intestinal gene expression [60–63]. Additionally, possibly growth-delayed or sterile outlier animals were removed from the dataset by analyzing the expression of spermatogenesis and oogenesis genes as described by Perez and colleagues [33]. For this, genes associated with Gene Ontology (GO) terms for oogenesis (n = 50 genes) and spermatogenesis (n = 57 genes) were used to identify outlier animals, and normalized read counts were summed per animal and plotted to identify outlier animals. Presence of outlier animals with very low levels of oogenesis and spermatogenesis gene expression was found to be independent of batch, condition, or genetic background. This resulted in a dataset consisting of 143 animals with gene expression data for a total of 2,638 genes for AID::knl-1 animals and 2,724 genes for AID::cdk-1 animals. The DEseq R package was used to perform differential gene expression analysis. Genes with an adjusted P value below 0.05 and a log2 fold change of gene expression within the top 25% (corresponding to an absolute log2 fold change above 0.514 in AID::knl-1 animals and 0.162 in AID::cdk-1 animals) were considered to be significantly and substantially differentially expressed. GO term enrichment analysis was performed using the InterMineR package for R. Fragments per kilobase of exon per million mapped fragments (FPKM) gene expression data from different developmental stages were retrieved from ModEncode through WormBase [64] and analyzed in R. A subset of intestinally expressed genes was composed by combining lists of genes that were identified as intestinally unique, enriched, or expressed in previous expression studies [60–63]. A Venn diagram was constructed using the BioVenn web application [65]. The RNA-sequencing data have been deposited at the Gene Expression Omnibus (GSE169330).

Lipid staining

Embryo lipid staining was performed as described previously [33]. In short, embryos were obtained by alkaline hypochlorite treatment of adult hermaphrodites, transferred to PCR tubes (Eppendorf, Hamburg, Germany) containing 4% formaldehyde in M9 buffer and subjected to 3 freeze-thaw cycles in liquid nitrogen and a 37°C water bath. Embryos were then incubated with 1 μg/ml BODIPY 493/503 (Thermo Fisher Scientific) in M9 buffer for 1 hour and washed 3 times with M9 + 0.01% Triton X-100 (Sigma #9001-93-1). Next, nuclei were stained with 5 ng/mL DAPI and imaged by epifluorescence microscopy.

Adults were fixed for lipid staining by incubating in 4% paraformaldehyde (Sigma-Aldrich 158127) for 1 hour at room temperature. Adults were then washed twice with PBS, resuspended in 95% ethanol and washed again in PBS. Next, fixed animals were incubated with 1 μg/ml BODIPY 493/503 (Thermo Fisher Scientific) in M9 buffer for 1 hour and washed 3 times with M9 + 0.01% Triton X-100. Animals were stained with 5 ng/mL DAPI and imaged by epifluorescence microscopy.

Viability assay

Worms were grown until the L4 stage under auxin or auxin-control conditions as described above. Single hermaphrodites were then transferred to new plates every 24 hours. On each day, the number of live progeny and unhatched eggs were counted 24 to 48 hours after removal of the adult worm to calculate the total brood size and embryonic lethality.

Progeny growth assay

Animals were grown until young adults under auxin, control, or no auxin condition as described above. Developmental transition timings were performed as previously described using a wash-off staging [33]. Briefly, plates containing synchronized gravid hermaphrodites were washed to remove all animals except embryos that remained attached to the solid media. To tightly synchronize a population of animals, larvae that had hatched within 1 hour after wash-off were collected and transferred to new plates. The fraction of animals that had undergone the transition from L3 to L4 was counted for 3 replicate plates containing around 150 worms each, in 3 separate experiments. To estimate the time at which half of the population had undergone the developmental transition, each population was scored at least twice, once before and once after half of the population had undergone the L4 transition. Count data were modeled using a binomial generalized linear model, and the time at which half of the population had undergone the developmental transition (T₅₀) was estimated using the function “dose.p” of the library “MASS” in R. Moreover, 95% confidence intervals were estimated by multiplying the standard error obtained by the “dose.p” function by 1.96. The significance of the difference between 2 calculated T₅₀ values was determined using the “comped” function in the library “drc” in R, which takes as input the 2 T₅₀ values, their standard error and a desired probability level (such as 0.95).

Single molecule fluorescence in situ hybridization (smFISH)

smFISH was performed as previously described [66] with some minor adjustments. In short, animals were fixed using 4% paraformaldehyde, resuspended in 70% ethanol, and stored at 4°C for up to 2 weeks. Short oligonucleotide probes complementary to the sfGFP and vit-2 sequences (see S3 Table for sequences) were designed using a web-based algorithm (www.biosearchtech.com/stellaris-designer) and ordered with a 3-amino modification to enable coupling to Cy5 (#PA25001; GE Healthcare, Chicago, IL, USA) as previously described [67]. Hybridization was performed overnight at 37°C in the dark, after which samples were washed and stained with DAPI. Z-stacks of Int3A and Int3V cells were generated using spinning disk microscopy and mRNA molecules were counted using the batch analysis mode in FISH-quant [68]. To quantify sfGFP mRNA densities, 4.5-μm substacks were made in Fiji of each cell centered around the nucleus/nuclei. From these substacks, mRNA spots were counted in 3 regions of 4.55 μm × 4.55 μm, and spot density was calculated as # of spots per μm³. To quantify the number of vit-2 mRNAs, substacks were made in Fiji of each cell, using the GFP-PH marker to determine cell boundaries. Cell outlines, as well as nuclei and nascent transcripts, were then annotated in FISH-quant and counted using the batch analysis mode. Nascent vit-2 transcripts were defined as nuclear smFISH spots that were at least 200 nm in XY and 300 nm in Z dimension and that contained at least 2 mRNAs, which was calculated by averaging the cytoplasmic mRNA spots and comparing the amplitude of the annotated transcription sites to individual mRNA molecules, using the transcription site quantification interface in FISH-quant.

Statistics and reproducibility

The sample size (n) as well as the number of replicate experiments performed for each experiment is described in the corresponding figure legend. All statistical analyses were performed using Prism version 8.4.3 (Graphpad, San Diego, CA, USA) except for sequencing data analyses. Quantitative data displayed as boxplots indicate the median and 25th to 75th percentile, error bars indicate min to max values, and individual values are shown as dots, unless indicated otherwise. Two-tailed unpaired Student t tests were used for pairwise comparisons between groups with similar Gaussian distributions. Mann–Whitney U tests were used for pairwise comparisons between groups with non-Gaussian distributions. The type of statistical test performed is indicated in the corresponding figure legends. Statistical significance values (P) are shown directly within the figure above pairwise comparisons.