The ciliary marginal zone of the zebrafish retina: clonal and time-lapse analysis of a continuously growing tissue

Abstract

Clonal analysis is helping us understand the dynamics of cell replacement in homeostatic adult tissues (Simons and Clevers, 2011). Such an analysis, however, has not yet been achieved for continuously growing adult tissues, but is essential if we wish to understand the architecture of adult organs. The retinas of lower vertebrates grow throughout life from retinal stem cells (RSCs) and retinal progenitor cells (RPCs) at the rim of the retina, called the ciliary marginal zone (CMZ). Here, we show that RSCs reside in a niche at the extreme periphery of the CMZ and divide asymmetrically along a radial (peripheral to central) axis, leaving one daughter in the peripheral RSC niche and the other more central where it becomes an RPC. We also show that RPCs of the CMZ have clonal sizes and compositions that are statistically similar to progenitor cells of the embryonic retina and fit the same stochastic model of proliferation. These results link embryonic and postembryonic cell behaviour, and help to explain the constancy of tissue architecture that has been generated over a lifetime.

Keywords: Ciliary marginal zone; Clonal analysis; Live imaging; Progenitor cells; Retina; Stem cells; Zebrafish.

Fig. 1.

Short-term clonal analysis. Rx2 functions as a marker for the CMZ of the zebrafish retina. (A) A maximum intensity projection of a transverse 12 µm section of a 5 dpf zebrafish retina, showing the expression of the transgenic reporter Tg(Rx2:GFPcaax). Rx2⁺ cells (green) are found in the CMZ as well as in two cell types of the neural retina, photoreceptors (PRs) and Müller glia. (B) High-magnification 1 µm section from a region of the CMZ (box in A), revealing that both RSCs (rings 1-2) and RPCs (rings 3-5) are Rx2⁺. Numbers indicate rings. Nuclei are labelled with DAPI (magenta). INL, inner nuclear layer; RGC, retinal ganglion cell layer. (C) Experimental procedure for labelling CMZ cells and capturing the short-term clonal composition. (D) Composition of short-term CMZ clones reveals the distinct proliferative fates in the CMZ population. (E) A single CMZ cell is photoconverted to red at 3 dpf (top left, indicated by arrow) and the resultant terminated clone at 5 dpf (top right, red) is shown with identified cell fates (RGC, retinal ganglion cell; PR, photoreceptors). Another example of a single CMZ cell photoconverted at 3 dpf (bottom left, indicated by arrow) and the resultant maintained clone observed at 5 dpf (bottom right, red). Note the dilution of the label in cells further from the lens, which is suggestive of more rapid division. In E, all Rx2⁺ cells have green membranes whereas the MAZe:Kaede cells have green cytoplasm that becomes red upon photoconversion. Scale bars: 10 µm.

Fig. 2.

Long-term polyclonal analysis. (A) Experimental procedure for labelling 5 dpf CMZ clones and following the maintaining CMZ polyclones at subsequent time points. (B) Photoconverted CMZ cells (red) at 5, 6, 8, 12 and 16 dpf, with the top two rows showing a maintained polyclone and bottom two rows a terminated polyclone at 16 dpf. Rx2⁺ cells have green membranes whereas the MAZe:Kaede cells have green cytoplasm that becomes red upon photoconversion. Yellow areas indicate colocalization of signals. Scale bar: 20 µm. The final panels in each series show the clonal geometry of each of these two clones. Black circles indicate ring 1; grey circles, ring 2; dark green to light green, rings 3-8. (C) Frontal view of an Rx2⁺ retina, with the cell ring position coded in colour (as shown in key). (D) Distribution of the position of the most peripheral cells in clones on the last observed day for all maintained clones (n=31) and the corresponding value for all terminated clones before termination (n=19). Error bars represents one sigma confidence interval of Poisson distribution.

Fig. 3.

Stabilization of CMZ clones is due to asymmetrical RSC division. (A) The rate of clone detachment events plotted as a function of age. Detachment rate is defined as the number of detachment cases (clones move away from rings 1-2 in a specific period of time) divided by the total number of cases surviving up to that time point, divided by the number of days in the time period. (B) Plot showing the average number of ring 1+ring 2 cells in the maintained clones as a function of age. Note that this number remains relatively constant in the maintained clones, indicating potential asymmetrical division of RSCs. (C) Model of RSC and RPC division, with α and β indicating the division rate of RSCs and RPCs, respectively.

Fig. 4.

Live imaging of cell division in the CMZ. (A) Schematic illustration of the two-photon imaging set-up. (B) 3D projection of one movie frame (t=69, z=16.5 µm), showing the CMZ region imaged. Arrows indicate areas of retinal pigment epithelium (RPE). Yellow rectangle represents the area shown in the following panels. The ring blood vessel is shaded light green. (C) Selected frames of a 4D time-lapse movie showing several cell divisions in a variety of orientation planes. Dividing cells are labelled with a coloured dot. A presumptive stem cell, dividing parallel to the edge of the retina close to the blood vessel, is labelled with a red asterisk. Each frame represents a maximum intensity projection (six frames, z=4.5 µm). Elapsed time shown in min:s. Dotted line indicates the edge of the blood vessel. Scale bar: 10 µm. (D) Computation of division angle. The 3D image stack is segmented into retina (blue) and blood vessel (green). Inset panel shows how the pixels belonging to the divisions are detected by analysing the difference between two consecutive or temporally close frames (depending on the time resolution of the imaging) and finding large contiguous volumes of statistically significant pixels in the difference image, followed by a manual validation step (see supplemental Materials and Methods). Once the divisions are detected, the normal of the division plane (yellow) is estimated and projected in the CMZ plane and the angle between the projected normal and the tangent to the CMZ circle at the closest point is computed. (E) Box plot of division angles detected in each ring of cells in the CMZ. Ring 1 cells divide more strictly axially compared with the random distribution in higher rings. Boxes span the first to the third quartile. Whiskers extend from the lowest value still within 1.5 interquartile range (IQR) of the lower quartile to the highest value still within 1.5 IQR of the upper quartile. (F) Estimated division rate in each ring of cells in the CMZ. Division rate is defined as the number of divisions observed divided by the number of cells imaged in the corresponding ring. Error bars represent one sigma confidence interval of Poisson distribution.

Fig. 5.

RPCs recapitulate the developmental history of central retina progenitors. (A) The frequency of fate combinations of sister cells in two-cell clones is comparable in the CMZ and the central retina. Colour in the matrix is normalized to the ratio of dominant fate combinations. Brightest red colour indicates the most common composition, and black means the fate combination does not exist. (B) Clonal composition of large CMZ clones (≥5 cells) resembles the composition of central retina clones induced at 24 hpf. None of the cell types except for RGCs shows a significant difference in the percentage of the clone (α=5%). (C) Schematic showing the theoretical model used to simulate the RPC clone size. Time-dependent mitosis and division mode probability in central retina development are plotted (adapted from He et al., 2012), with the CMZ progenitors randomly selected from the pool of progenitors in the first 17 h within the lineage, indicated in the purple box. (D) An example of the RPC clone in the CMZ recapitulating the age distribution within a prototypical cell lineage in a central retina progenitor. I₀, initial progenitor; I₁, primary progenitor; I₂, secondary progenitor; I₃, tertiary progenitor, all residing in the CMZ region (green); D, differentiated cells residing in the central retina (blue). (E) Fit between model predictions (red line with shaded pink region showing 95% plausible intervals due to finite sampling) and size distribution (blue crosses) of RPC clones. In this figure, data shown in the right panel of A, the red bars in B and the model shown in C have all been extracted from He et al. (2012).