1. Developmental Biology
Download icon

mTORC1-induced retinal progenitor cell overproliferation leads to accelerated mitotic aging and degeneration of descendent Müller glia

  1. Soyeon Lim
  2. You-Joung Kim
  3. Sooyeon Park
  4. Ji-heon Choi
  5. Young Hoon Sung
  6. Katsuhiko Nishimori
  7. Zbynek Kozmik
  8. Han-Woong Lee
  9. Jin Woo Kim  Is a corresponding author
  1. Department of Biological Sciences, Korea Advanced Institute of Science and Technology, Republic of Korea
  2. Department of Biochemistry, College of Life Science and Biotechnology, Yonsei University, Republic of Korea
  3. Department of Convergence Medicine, Asan Medical Center, University of Ulsan College of Medicine, Republic of Korea
  4. Department of Obesity and Internal Inflammation; Bioregulation and Pharmacological Medicine, Fukushima Medical University, Japan
  5. Institute of Molecular Genetics of the Czech Academy of Sciences, Czech Republic
Research Article
  • Cited 0
  • Views 415
  • Annotations
Cite this article as: eLife 2021;10:e70079 doi: 10.7554/eLife.70079

Abstract

Retinal progenitor cells (RPCs) divide in limited numbers to generate the cells comprising vertebrate retina. The molecular mechanism that leads RPC to the division limit, however, remains elusive. Here, we find that the hyperactivation of mechanistic target of rapamycin complex 1 (mTORC1) in an RPC subset by deletion of tuberous sclerosis complex 1 (Tsc1) makes the RPCs arrive at the division limit precociously and produce Müller glia (MG) that degenerate from senescence-associated cell death. We further show the hyperproliferation of Tsc1-deficient RPCs and the degeneration of MG in the mouse retina disappear by concomitant deletion of hypoxia-induced factor 1-alpha (Hif1a), which induces glycolytic gene expression to support mTORC1-induced RPC proliferation. Collectively, our results suggest that, by having mTORC1 constitutively active, an RPC divides and exhausts mitotic capacity faster than neighboring RPCs, and thus produces retinal cells that degenerate with aging-related changes.

Editor's evaluation

Using a broad range of genetic and biochemical strategies, this study shows that hyperactivation of mTOR in retinal progenitors induces hyperproliferation and thus prematurely exhaust their mitotic capacity. They also show that these effects are related to Hif1alpha activation and metabolism. The study will be of considerably interest beyond field of retinal development, as it clearly links cell metabolism to tissue growth and perhaps competition.

https://doi.org/10.7554/eLife.70079.sa0

Introduction

Neural progenitor cells (NPCs) divide repeatedly during development to generate the cells of vertebrate neural tissues (Homem et al., 2015; Obernier and Alvarez-Buylla, 2019). NPCs are present for a limited period before entering a final cell division for the differentiation to neurons and glia. Consequently, NPCs are not seen in the majority of adult neural tissues, except for sub-brain areas that exhibit continued neurogenesis.

The factors that determine when an NPC enters the final division and how long it keeps division capacity, however, still remain elusive. It has been identified that the mitotic characteristics of NPCs change continually in the development of neural tissues. Most NPCs divide symmetrically to expand themselves during early development; thereafter, the asymmetrically dividing NPCs increase to preserve the NPC population while actively generating the various types of neurons and glia (Homem et al., 2015; Obernier and Alvarez-Buylla, 2019). The length of the NPC cell cycle also increases with development, while the NPC division capacity decreases (Alexiades and Cepko, 1996; Ohnuma and Harris, 2003). Given the heterogeneity of the NPC division mode and cell cycle length, it is believed that cumulative division numbers might be diversified among NPCs even within the same neural tissues in development. Consequently, some NPCs might have already reached their mitotic division limits while neighboring NPCs are still able to divide further. However, the division capacity of an NPC in developing neural tissue has not been empirically quantified to date.

The mouse retina has been used as a model system in studies aiming to identify general features of mammalian NPC proliferation and differentiation (Cepko, 2014). Two retinal progenitor cell (RPC) populations have been identified in mouse retina (Clark et al., 2019). The early RPC population produces retinal ganglion cells (RGCs), amacrine cells (ACs), horizontal cells (HCs), cone photoreceptors (cPRs), and some rod photoreceptors (rPRs), whereas the late RPC population mainly produces bipolar cells (BCs), Müller glia (MG), and some rPRs (Cepko, 2014; Clark et al., 2019). Given the temporal-specific development of retinal cell types, cumulative division numbers of RPCs producing MG, which is the last-born retinal cell type, are likely different from those of RPCs generating RGCs, which is the first-born retinal cell type.

The decision of an RPC to exit the cell cycle and undergo differentiation is regulated by various factors. For example, RPCs extended division capacity in mice deficient of the cell cycle-dependent kinase inhibitor 1b (Cdkn1b/p27) (Levine et al., 2000), whereas they prematurely exited the cell cycle and were extinguished precociously in mice lacking cyclin D1 (Ccnd1) (Das et al., 2012). Consequently, the production of MG is decreased in Ccnd1-deficient mouse retina, whereas it is extended in Cdkn1-deficient mouse retina. Therefore, cumulative division numbers of Cdkn1-deficient RPCs are likely to be greater than those of Ccnd1-deficient RPCs when these cells undertake the production of MG in the postnatal retina.

The speed of the cell cycle could also affect the cumulative division number of RPCs as well. Tuberous sclerosis complex 1 (Tsc1)-deficient mouse RPCs, which have hyperactive mechanistic target of rapamycin complex 1 (mTORC1), were found to complete a cell cycle more quickly than wild-type mouse RPCs (Choi et al., 2018). This resulted in faster accumulation of newborn cells in the Tsc1-deficient mouse retinas than in wild-type retinas. Here, we also found faster accumulation of cells in the mouse retina where Tsc1 was deleted in the minor RPC population derived from the ciliary margin (CM). This made the CM RPC-derived Tsc1-deficient clones invade into neighboring areas, where wild-type clones grow slowly, and occupy almost entire territory of the mature mouse retina. However, later, the retinal cells, especially the MG, produced from Tsc1-deficient CM RPCs take on senescent characteristics and degenerate to form rosette structures in the retinas. The Tsc1-deficient retinal cells were found to be mitotically older than those derived from neighboring wild-type RPC clones. The precocious aging phenotypes of the Tsc1 conditional knock-out (cko) mouse retina were rescued by concomitant deletion of hypoxia-induced factor 1-alpha (Hif1a), which supports RPC proliferation by inducing the expression of glycolytic enzymes that can supply ATP in Tsc1-deficient RPCs. These results suggest that there are limits to RPC mitotic division that can be reached precociously by hyperexpansion of an RPC clone in the developing retina.

Results

Degeneration of MG derived from Tsc1-deficient CM RPCs

mTORC1 activity, which leads to the phosphorylation of ribosomal protein S6 (pS6) via the activation of S6 kinase 1 (S6K1), was detectable in developing mouse retina but was absent in the neighboring retinal pigment epithelium (RPE) and CM (Figure 1—figure supplement 1). However, it is unknown why mTORC1 activity is diversified in the optic neuroepithelial continuum of the retina-CM-RPE, which were shown to exhibit differential proliferation rates (Moon et al., 2018). To understand the physiological importance of the spatially differentiated mTORC1 activation, we ectopically increased mTORC1 activity in the RPE and CM by deleting Tsc1, which is a negative upstream regulator of mTORC1 (Gao et al., 2002; Saxton and Sabatini, 2017). To this end, we bred Tsc1flox/flox (Tsc1fl/fl) mice with tyrosinase-related protein 1-Cre (Tyrp1-Cre) mice, which express Cre recombinase in both the RPE and the CM cells that have a potential to become RPCs in the peripheral retina (Fischer et al., 2013; Mori et al., 2002; Figure 1A). As expected, Cre-affected cells, which were visualized by a fluorescent Cre reporter ROSA26EYFP (R26EYFP), were detectable in the RPE, CM, and peripheral retina of Tsc1fl/+;Tyrp1-Cre mice (Figure 1B, top row). The R26EYFP-positive cells were found farther in the central part of Tsc1fl/fl;Tyrp1-Cre mouse retinas, suggesting that the CM-derived Tsc1-deficient cell populations were expanded more centrally in the retinas.

Figure 1 with 4 supplements see all
MG degeneration and rosette formation in Tsc1fl/fl;Tyrp1-Cre mouse retina.

(A) Developmental changes of eye and retinal structures in Tsc1fl/+;Tyrp1-Cre and Tsc1fl/fl;Tyrp1-Cre littermate mice were examined by hematoxylin and eosin (H&E) staining of the eye sections. ONL, outer nuclear layer; OPL, outer plexiform layer; INL, inner nuclear layer; IPL, inner plexiform layer; GCL, ganglion cell layer. (B) The Cre-affected cells in the mouse retinas were visualized by R26EYFP Cre reporter, and mTORC1 activation of the cells was determined by immunostaining of pS6. Nuclei of the cells were visualized by DAPI staining. (C) Distributions of MG in the mouse retinas were examined by immunostaining of the MG markers, Sox9 and glutamine synthetase (GS). (D) Retinal cell type-specific marker-positive cells among DAPI-positive total retinal cells in 350 μm × 350 μm areas were counted and shown their relative values against those of Tsc1fl/+;Tyrp1-Cre mouse retinas in the graph. Representative staining images of retinal markers are provided in (C) and Figure 1—figure supplement 2A. Error bars denote standard deviations (SD). The numbers of samples are 5 from five independent litters. **p < 0.01; ***p < 0.001; ****p < 0.0001.

We found that the eyes of Tsc1fl/fl;Tyrp1-Cre mice at P14 and P30 were significantly smaller than those of Tsc1fl/+;Tyrp1-Cre littermates (Figure 1A and B, top rows), and their ciliary body (CB) and iris were malformed (arrows in Figure 1A, middle row). Moreover, the Tsc1fl/fl;Tyrp1-Cre mouse retinas exhibited multiple rosette structures (arrowheads in Figure 1A, bottom row). However, these phenotypes were not observed in P7 Tsc1fl/+;Tyrp1-Cre and Tsc1fl/fl;Tyrp1-Cre littermate mouse eyes (Figure 1A and B, two leftmost columns). The results suggest that the structural alterations in Tsc1fl/fl;Tyrp1-Cre mouse retinas began after the first postnatal week.

It is known that the loss of MG is a major cause of retinal rosette formation (Willbold et al., 2000). We found that the cells expressing the MG markers, including p27, SRY-box transcription factors 2 and 9 (Sox2 and Sox9), and glutamine synthetase, were decreased significantly in P14 and P30 Tsc1fl/fl;Tyrp1-Cre mouse retinas compared with Tsc1fl/+;Tyrp1-Cre littermate retinas, whereas the numbers of other retinal cell types were not significantly different in those two retinas (Figure 1C and D; Figure 1—figure supplement 2A and B). However, the numbers of MG were rather increased in P7 Tsc1fl/fl;Tyrp1-Cre mouse retina (Figure 1D; Figure 1—figure supplement 2B); this is likely due to the developmental acceleration of MG production, as it was reported previously (Choi et al., 2018). The numbers of apoptotic cells, which were assessed by TUNEL (terminal deoxynucleotidyl transferase dUTP nick end labeling) and immunostaining of cleaved active caspase-3 (Casp-3), were also elevated in the MG population of Tsc1fl/fl;Tyrp1-Cre mouse retinas (Figure 1—figure supplement 3A,B,C,D,E,F,G). Furthermore, the apoptotic cells were enriched in the Tsc1-deficient cell population positive for the Cre reporter, ROSA26tdTomato (R26tdTom) (Figure 1—figure supplement 3A and C), suggesting that Tsc1 deletion had autonomous effects on MG degeneration. These findings suggest that the number of MG in Tsc1fl/fl;Tyrp1-Cre mice might decrease below a critical level needed to maintain an intact retinal structure, due to the enhanced degeneration of MG.

Retinal rosettes and degeneration of MG were also observed in Tsc2fl/fl;Tyrp1-Cre mice, in which the other TSC component, Tsc2 (Inoki et al., 2002; Saxton and Sabatini, 2017), was deleted in the RPE and CM cells (Figure 1—figure supplement 4A and C). However, eyes of Tsc2fl/fl;Tyrp1-Cre mice were not significantly different in size compared to those of Tsc2fl/+;Tyrp1-Cre mice, and their CB and iris were intact (Figure 1—figure supplement 4A and B). These results suggest that the microphthalmia and CB/iris malformation of Tsc1fl/fl;Tyrp1-Cre mice is not due to mTORC1 hyperactivation; instead, they might be caused by cellular events that are regulated by Tsc1.

MG degeneration and retinal rosette formation are not caused by the hyperactivation of mTORC1 in mature retina

To examine whether the MG degeneration and retinal rosette formation were due to mTORC1 hyperactivation in MG, we deleted Tsc1 using the MG-specific solute carrier family 1 member 3-CreERT2 (Slc1a3-CreERT2), which is active in the presence of estrogen analog, tamoxifen (Tam) (de Melo et al., 2012). However, we could not observe MG degeneration or retinal rosettes in P30 Tsc1fl/fl;Slc1a3-CreERT2 mice, from which Tsc1 was deleted in MG beginning at P10 by Tam injection (Figure 2A).

Inhibition of mTORC1 cannot suppress MG degeneration in the mature mouse retina.

(A) The Tsc1fl/+;Slc1a3-CreERT2 and Tsc1fl/fl;Slc1a3-CreERT2 littermate mice were injected with tamoxifen (Tam) at P10 before the isolation of the eyes for cryosection at P30. The eye sections were stained with hematoxylin and eosin (H&E) to investigate the eye and retinal structures. Distributions of the cells expressing R26tdTom Cre reporter, mTORC1 activitation marker pS6, and MG marker Sox9 in the eye sections were examined by immunostaining. Nuclei of the cells in the sections were visualized by DAPI staining. (B) Tsc1fl/+;Tyrp1-Cre and Tsc1fl/fl;Tyrp1-Cre littermate mice were injected with rapamycin (2 mg/kg) daily from P7 to P13 to inhibit mTORC1. Alternatively, the mice were injected with same volume of the vehicle (5% poly-ethylene glycol and 5% Tween 80 in PBS). Retinal structures of the injected mice were investigated at P14 by H&E staining of their eye sections. Distribution of Cre-affected cells and mTORC1 activation of the cells were examined by co-immunostaining of R26EYFP and pS6. Distributions of MG in the mouse retinas were examined by immunostaining of Sox9.

We also found that the retinal rosettes of Tsc1fl/fl;Tyrp1-Cre mice were still evident when mTORC1 was inhibited by daily injection of a chemical inhibitor, rapamycin, during the second postnatal week (Figure 2B). The numbers of MG in Tsc1fl/fl;Tyrp1-Cre mouse retinas were also not recovered by rapamycin treatment, whereas pS6 was completely absent from the rapamycin-treated mouse retinas. Collectively, these results suggest that the degenerative phenotypes are not due to mTORC1 activation in the MG; instead, they likely arise from mTORC1 activation in the developing mouse retina.

MG degeneration and retinal rosette formation cannot be induced by Tsc1 deletion in the majority of RPC and RPE populations

Interestingly, in contrast to Tsc1fl/fl;Tyrp1-Cre mice, retinal rosettes were not seen in Tsc1fl/fl;Chx10-Cre and Tsc1fl/fl;Mlana-Cre mice (Figure 3A, third and fifth columns from left), in which Tsc1 was deleted from most of the RPCs by Chx10-EGFP/cre (Chx10-Cre) and from the RPE by Mlana-Cre, respectively (Aydin and Beermann, 2011; Rowan and Cepko, 2004). The embryonic and early postnatal mouse RPE was shown to make direct contact with adjacent RPCs to regulate neurogenic capacity of the RPCs (Ha et al., 2017). Therefore, we speculated that the phenotypes of Tsc1fl/fl;Tyrp1-Cre mice might be induced only when Tsc1 was lost commonly in the RPE and RPC. However, we failed to find the decrease of MG as well as retinal rosettes in Tsc1fl/fl;Chx10-Cre;Mlana-Cre mouse eyes, in which Tsc1 was deleted in RPC and RPE together (Figure 3A [rightmost column] and 3B). The numbers of RPE were also unchanged in the eyes of the double Cre-expressing mice (Figure 3A and C).

Normal eye and retinal structures in the mice lacking tuberous sclerosis complex 1 (Tsc1) in majority retina and retinal pigment epithelium (RPE) populations.

(A) Retinal structures of P30 mice deleted of Tsc1 in majority retinal progenitor cells (RPCs) by Chx10-Cre or in the RPE by Mlana-Cre were investigated by hematoxylin and eosin (H&E) staining of the eye sections. Distribution of Cre-affected cells and mTORC1 activitation of the cells were examined by co-immunostaining of R26EYFP and pS6. Distributions of MG in the mouse retinas were examined by immunostaining of Sox9. RPE in whole-mount eye cups was visualized by immunostaining of Otx2, which locates in the RPE nuclei, and F-actin, which marks RPE cell boundary. Numbers of Sox9-positive MG in the retinal sections (B) and Otx2-postive RPE in the whole-mount eye cups (C) were counted and shown in the graphs. Error bars denote SD and numbers of samples are 6 from four independent litters.

Given the absence of Cre activity in the CM area of Tsc1fl/fl;Chx10-Cre;Mlana-Cre mice (see R26EYFP Cre reporter signals in Figure 3A), we then questioned whether the MG degeneration and rosette formation could be caused by Tsc1 deletion in the CM population. To explore this, we deleted Tsc1 from RPCs in the peripheral retina and the inner CM cell population using Pax6-cre,GFP (Pax6-aCre) (Marquardt et al., 2001). However, the eyes and retinas appeared normal in Tsc1fl/fl;Pax6-aCre mice, and their MG cell numbers were not greatly different from those in Tsc1fl/+;Pax6-aCre littermate mice (Figure 4A, third and fourth columns from left). Given the absence of Pax6-aCre activity in the pigmented outer CM cells, we next deleted Tsc1 from the entire optic neuroepithelia-derived cell population, including the retina, the inner and outer CM, and the RPE, using Rax-cre (Klimova et al., 2013), and tested whether Tsc1 deletion in the outer CM population is necessary for the observed phenotypes. However, Tsc1fl/fl;Rax-Cre mice also exhibited normal retinal morphologies and MG cell numbers (Figure 4A, rightmost column).

Figure 4 with 4 supplements see all
MG degeneration is caused by clonal hyperexpansion of tuberous sclerosis complex 1 (Tsc1)-deficient cells in developing mouse retina.

(A) Eye and retinal structures of P30 mice with the indicated genotypes were examined by hematoxylin and eosin (H&E) staining. Distribution of Cre-affected cells and mTORC1 activation of the cells were examined by co-immunostaining of R26tdTom and pS6. (B) Cre-affected Tsc1-heterozygote (Tsc1fl/+;Cre) and Tsc1-deficient (Tsc1fl/fl;Cre) cells, which emitted red fluorescence of R26tdTom Cre reporter, are isolated from Cre-unaffected wild-type cells in same retinas by FACS and the histograms are presented. (C) Relative composition of R26tdTom-positive and -negative cell populations in the retinas are shown in the graph. Error bars denote SD. Numbers of samples analyzed are shown in the graph columns. ***p < 0.001; n.s., not significant. (D) Distributions of the Cre-affected cells in the corresponding E14.5 and P30 mouse retinas are summarized in the drawings. (E) Developmental changes of R26tdTom-positive Cre-affected population in Tsc1fl/+;Tyrp1-Cre and Tsc1fl/fl;Tyrp1-Cre littermate mouse retinas were determined by FACS and shown in the graph. Error bars denote SD. Numbers of samples analyzed are shown in the graph (four independent litters). **p < 0.01; ****p < 0.0001. (F) Tsc1+/- and Tsc1-/- retinal progenitor cells (RPCs) were isolated from E13 Tsc1fl/+;Chx10-Cre and Tsc1fl/fl;Chx10-Cre and were cultured to form the neurospheres. Representative images of the neurospheres at the indicated post-culture days are provided. (G) Sizes of neurospheres in culture were counted at the indicated post-culture days and their average values are shown in the graph. (H) Average numbers of neurosphere in the indicated area are also shown in the graph. Numbers of samples analyzed are shown in the graphs (± independent culture batches). *p < 0.05; ***p < 0.001; ****p < 0.0001; n.s., not significant.

MG degeneration is correlated with the clonal expansion of Tsc1-deficient RPCs in the retina

The Tsc1-deficient RPCs can complete a cell cycle faster than wild-type RPCs (Choi et al., 2018), therefore the time necessary to double the population is likely shorter for a Tsc1-deficient RPC than for a wild-type RPC in the same retina. This, therefore, could enable the small Tsc1-deficient population in the peripheral retina to expand into the central retina where wild-type populations expand slowly in Tsc1fl/fl;Tyrp1-Cre mice (Figure 1B, top row). However, a Tsc1-deficient RPC in Tsc1fl/fl;Rax-Cre and Tsc1fl/fl;Chx10-Cre mouse retinas might not have this competitive advantage over their neighboring RPCs for clonal expansion, since the neighboring RPCs also lose Tsc1 and expand as quickly as the Tsc1-deficient RPC.

To confirm the differential clonal expansion of Tsc1-deficient cells in each Tsc1-cko mouse retina, we collected R26tdTom-positive Cre-affected cells, which are potential Tsc1-heterozygous and Tsc1-deficient cells in Tsc1fl/+;Cre and Tsc1fl/fl;Cre mouse retinas, respectively, using fluorescence-activated cell sorting (FACS). Our FACS results showed that the percentages of retinal cells affected by the Cre recombinases in Tsc1fl/+ mouse retinas were approximately 48% for Tyrp1-cre, 65% for Pax6-aCre, 81% for Chx10-Cre, and 92% by Rax-Cre. The cells were increased to about 85% in Tsc1fl/fl;Tyrp1-Cre mouse retinas; 89% in Tsc1fl/fl;Pax6-aCre mouse retinas; 93% in Tsc1fl/fl;Chx10-Cre mouse retinas; and 95% in Tsc1fl/fl;Rax-Cre mouse retinas (Figure 4B and C). This resulted in 1.77-, 1.36-, 1.16-, and 1.03-fold increases, respectively, of Cre-affected cells in the Tsc1fl/fl;Cre mouse retinas comparing with their littermate Tsc1fl/+;Cre mouse retinas (Figure 4C). These results can also be translated to overproduction of Tsc1-deficient cells by 77%, 35.9%, 16.3%, and 3.2% relative to the levels expected based on the penetrance of each Cre driver. This clonal expansion had begun during the embryonic stages and continued until P14, when retinal histogenesis was completed (Figure 4D and E; Figure 4—figure supplement 1). These results suggest the clonal expansion power of a Tsc1-deficient RPC is inversely correlated with the population of Tsc1-deficient RPCs in developing mouse retina.

Further, we deleted Tsc1 in smaller retinal population than those affected by Tyrp1-Cre using Tyrp1-CreERT2, which can be activated in the CM and RPE subpopulation by Tam. As we expected, the R26EYFP-positive CreER-affected cells were detected sparsely in P30 Tyrp1-CreERT2 mouse retina (Figure 1A), when Tam was injected by E9.5 (Figure 4—figure supplement 2A). We could also find retinal rosettes in the Tsc1f/f;Tyrp1-CreERT2 mice, which were injected with Tam at E9.5, but not in those injected with Tam at P0 (Figure 4—figure supplement 2B). The rosette areas showed the decrease of MG and the increase of apoptotic cells (Figure 4—figure supplement 2C–E ). These results suggest the degeneration of MG derived from Tsc1-deficient CM RPCs given rise from early mouse embryos, which have greater division capacity than those from later stages.

Earlier arrival of Tsc1-deficent mouse RPCs at the division limit

Given the positive relationship between the clonal expansion power of Tsc1-deficient cells and MG degeneration in the Tsc1-cko mouse retinas, we hypothesized that the Tsc1-deficient RPCs in Tsc1fl/fl;Tyrp1-Cre mouse retinas were likely to be too old, in mitotic terms, to preserve the survival program intact in their descendent MG. In the other words, the Tsc1-deficient RPCs were close to or had exceeded their mitotic division limit when they produced MG in Tsc1fl/fl;Tyrp1-Cre mouse retinas, whereas the RPCs in the other Tsc1-cko mice had not yet reached that limit.

To test this hypothesis, we compared the division capacities of Tsc1-heterozygote and Tsc1-deficient mouse RPCs in vitro. The RPCs were isolated from E13 Tsc1fl/+;Chx10-Cre and Tsc1fl/fl;Chx10-Cre littermate mouse embryo retinas, respectively, and cultured to form neurospheres. We found that Tsc1-deficient RPCs divided faster to form larger neurospheres than those derived from Tsc1-heterozygote RPCs at one post-culture week (Figure 4F). However, the Tsc1-deficient neurospheres stopped expanding and the cells comprising the neurospheres started to degenerate after 3 post-culture weeks, while the Tsc1-heterozygote neurospheres expanded slowly and maintained in relative intact forms by 5 post-culture weeks (Figure 4F and G). Consequently, the numbers of Tsc1-deficient neurospheres became less than those of Tsc1-heterozygote neurospheres after 4 post-culture weeks (Figure 4F and H). The fast expansion and precocious degeneration of Tsc1-deficient neurospheres were not affected by the presence of wild-type neurospheres in the neighbor, suggesting that overproliferation is an intrinsic property of Tsc1-deficient RPCs (Figure 4—figure supplement 3A – D). These results suggest that the RPC has a limited division capacity that can be reachable earlier when the cells divide faster, as seen in the Tsc1-deficient RPCs (Figure 4—figure supplement 4).

Accelerated mitotic aging of Tsc1-deficient mouse RPCs

To further validate the idea that the Tsc1-deficient RPCs in Tsc1fl/fl;Tyrp1-Cre mice overproliferated to produce MG after exceeding their division limits, we explored the relative division numbers of retinal cells comprising various Tsc1-cko mouse retinas. Telomeric DNA sequences are shortened after every cell division unless they are recovered by telomerase (Calado and Dumitriu, 2013; Shay, 2016). Given the absence of telomerase reverse transcriptase (Tert) and telomerase RNA component (Terc) mRNA expression in embryonic and postnatal mouse retinas (data not shown), we expected that the telomeric sequences would be shortened constantly in mouse RPCs after each division. In support of this, we found that the average telomere length in mouse retina cells decreased with age (Figure 5—figure supplement 1). We further found that the telomere shortening was faster in Tsc1fl/fl;Tyrp1-Cre mouse retinas compared to Tsc1fl/+;Tyrp1-Cre littermate mouse retinas (Figure 5A; Figure 5—figure supplement 1). However, there was no significant difference in telomere length between other Tsc1fl/+;Cre and Tsc1fl/fl;Cre littermate mouse retinas (Figure 5A). These results suggest the ability of Tsc1-deficient RPCs to overproliferate is inversely related with their frequencies in the retina.

Figure 5 with 4 supplements see all
Mitotic aging and senescence of tuberous sclerosis complex 1 (Tsc1)-deficient retinal cells.

(A) Telomere lengths of the retinal cells, which were isolated from P30 Tsc1fl/+ and Tsc1fl/fl mice with the indicated Cre drivers, were compared with that of control sample, and relative values are shown in the graph. Error bars denote SD. Numbers of samples analyzed are shown in the graph. All samples were obtained from independent litters. **p < 0.01; n.s., not significant. (B) Telomere lengths of the FACS-isolated retinal cells from P14 Tsc1fl/+;Tyrp1-Cre and Tsc1fl/fl;Tyrp1-Cre littermate mice were compared with that of control sample, and their relative values are shown in the graph. *p < 0.05; ***p < 0.001. (C) Senescence markers (β-gal, p53, p21, and c-Myc) expressed in the corresponding P30 Tsc1fl/+;Cre and Tsc1fl/fl;Cre mouse retinas were detected by Western blot (WB). Relative amounts of the proteins used in each sample were determined by WB detection of β-actin. (D) Expression of senescence markers (β-gal and p53) in the MG was determined by co-immunostaining with an MG marker, Sox9. The cells experienced Cre-mediated recombination were also visualized by R26tdTom reporter. Images in the right columns are the magnified versions of the boxed areas indicated by corresponding numbers in the low magnification images. (E) Sox9-positive MG, Pax6-positive amacrine cell (AC), and Vsx2-positive bipolar cell (BC) populations that express the senescence markers were determined and shown in the graph (Pax6 and Vsx2 staining images are provided in Figure 5—figure supplement 4). Number of samples analyzed are shown in the graph (five independent litters). n.d., not detected. (F) Eye and retinal structures of P30 Rptorfl/+;Tyrp1-Cre and Rptorfl/fl;Tyrp1-Cre mice were investigated by hematoxylin and eosin (H&E) staining of the eye sections. Distribution of the Cre-affected cells and mTORC1 activation of the cells were examined by the immunostaining of R26tdTom and pS6, respectively. (G) R26tdTom-positive Cre-affected cell population in the mouse retinas were quantified by FACS and shown in the graph. ****p < 0.0001. (H) Telomere lengths of unsorted (whole retina) and the FACS-isolated retinal cells from P14 Rptorfl/+;Tyrp1-Cre and Rptorfl/fl;Tyrp1-Cre mice were compared with that of control sample and their relative values are shown in the graph. Number of samples analyzed are shown in the graph. *p < 0.05. (I) Relative levels of senescence markers and mTORC1 pathway components of the mouse retinas were analyzed by WB. (J) Distribution of the cells expressing senescence markers and R26tdTom Cre reporter in the retinas was also examined by immunostaining.

To examine whether telomere shortening was an autonomous event of Tsc1-deficient cells, we next compared the telomere lengths of FACS-isolated wild-type and Tsc1-deficent cells in the same Tsc1fl/fl;Tyrp1-Cre mouse retinas. We found that the telomeres in R26tdTom-positive Tsc1-deficient cells of P14 Tsc1fl/fl;Tyrp1-Cre mouse retinas were significantly shorter than those in R26tdTom-negative neighboring wild-type cells (Figure 5B). The accelerated telomere shortening was also observed in the neurospheres derived from Tsc1-deficent RPCs, which were cultured homogenously or together with wild-type neurospheres (Figure 4—figure supplement 3F). These results suggest that Tsc1 deletion had autonomous effects on telomere shortening.

Senescence-associated cell death of MG in hyperexpanded Tsc1-deficient retinal clones

Extensive telomere shortening can cause mitotic catastrophe and subsequent cell death (Calado and Dumitriu, 2013; Shay, 2016), suggesting that the degeneration of MG in Tsc1fl/fl;Tyrp1-Cre mouse retina could be caused by telomere shortening beyond a critical length. To test this possibility, we bred Tsc1fl/fl;Tyrp1-Cre mice with Cre-d-Tert (dTert) transgenic mice, which express mouse Tert cDNA in Cre-affected cells (Hidema et al., 2016), to recover telomeres in Tsc1-deficient retinal cells (Figure 5—figure supplement 2C). However, MG degeneration and retinal rosettes were still observed in Tsc1fl/fl;Tyrp1-cre;dTert mouse retinas (Figure 5—figure supplement 2A and B). This suggests that telomere shortening is unlikely a direct cause of MG degeneration in Tsc1fl/fl;Tyrp1-Cre mice.

The more cells divide the more damage accumulates, resulting in senescence-associated cell death (Shay, 2016). We found that markers of senescence, including β-galactosidase (β-gal), transformation-related protein 53 (Trp53, p53), p21 cyclin-dependent kinase inhibitor 1 A (p21Cip1, p21), and c-Myc (Kuilman et al., 2010), were detected in P30 Tsc1fl/fl;Tyrp1-Cre mouse retinas but not in Tsc1fl/+;Tyrp1-Cre littermate mouse retinas (Figure 5C; Figure 5—figure supplement 3A). Furthermore, β-gal and p53 were enriched more in R26tdTom-positive Tsc1-deficient MG (Figure 5; Figure 5—figure supplement 4), which degenerated to form the retinal rosettes (Figure 1C,D), than in R26tdTomato-positive Tsc1-deficient AC (Pax6-positive) and BC (Vsx2-positive) populations, which remained intact in the retinas (Figure 1D; Figure 1—figure supplement 2A). The senescence markers were also increased in Tsc2fl/fl;Tyrp1-Cre mouse retinas (Figure 5—figure supplement 3B), suggesting that the retinal senescence occurred in an mTORC1-dependent manner. In contrast, the senescence markers were not detectable in Tsc1fl/fl;Rax-Cre, Tsc1fl/fl;Chx10-Cre, and Tsc1fl/fl;Pax6-aCre mouse retinas (Figure 5C), suggesting a positive relationship between RPC clonal expansion power and senescence.

In contrast to the accelerated cell cycle progression of Tsc1-deficient RPCs, mouse RPCs lacking Rptor (regulatory-associated protein of mTOR), a key component of mTORC1 (Kim et al., 2002), were found to exit the cell cycle precociously (Choi et al., 2018). Consequently, the Rptor-deficient RPCs failed to expand, resulting in the compression of R26tdTom-positive Rptor-deficient clones in Rptorfl/fl;Tyrp1-Cre mouse retinas compared to those in Rptorfl/+;Tyrp1-Cre littermate mouse retinas (Figure 5F and G). These results suggest that retinal cells derived from the Rptor-deficient RPCs might be mitotically younger than neighboring wild-type cells in the Rptorfl/fl;Tyrp1-Cre mouse retinas.

Thus, we compared the relative telomere lengths of FACS-isolated Rptor-deficient cells and wild-type cells in the same retinas. We found that R26tdTom-positive Rptor-deficient cells of P14 Rptorfl/fl;Tyrp1-Cre mouse retinas had longer telomeres than R26tdTom-negative wild-type neighbors and R26tdTom-positive Rptor-heterozygote cells of Rptorfl/+;Tyrp1-Cre littermate mouse retinas (Figure 5H). However, we did not observe a significant difference in the telomere lengths of R26tdTom-negative wild-type cells in Rptorfl/fl;Tyrp1-Cre and Rptorfl/+;Tyrp1-Cre littermate mouse retinas, suggesting that there was no compensatory overproliferation of wild-type RPCs after the depletion of Rptor-deficient RPCs in Rptorfl/fl;Tyrp1-Cre mouse retinas. There was also no increase in the senescence markers in Rptorfl/fl;Tyrp1-Cre mouse retinas, either (Figure 5I and J). Together, our results suggest that senescence occurs in retinas where RPCs have exceeded their division limits, as seen in Tsc1fl/fl;Tyrp1-Cre mice, but not in retinas where RPCs have not reached the limits, as seen in the other Tsc1-cko and Rptorfl/fl;Tyrp1-Cre mice.

Hif1a is necessary for the hyperexpansion of Tsc1-deficient retinal clones

We next sought for the strategies to suppress the RPC overproliferation that led to the clonal hyperexpansion of Tsc1-deficient cells and consequent degeneration of MG in Tsc1fl/fl;Tyrp1-Cre mouse retinas. We previously showed that mTORC1 induces immunoproteasomes to increase the turnover of cyclins and thereby accelerate RPC cell cycle progression (Choi et al., 2018). We thus deleted the three catalytic subunit genes, proteasome subunit beta 8, 9, and 10 (Psmb8,9,10), to eliminate the contribution of immunoproteasomes to the mTORC1-induced mitotic acceleration of Tsc1-deficient mouse RPCs (Figure 6—figure supplements 1 and 2B). We found that retinal rosettes were absent from approximately a quarter of in Psmb8-/-,9-/-,10-/- (Psmb-tko);Tsc1fl/fl;Tyrp1-Cre mouse retinas, and R26tdTom-positive Tsc1-deficient retinal clone size was decreased in comparison with that of Tsc1fl/fl;Tyrp1-Cre mouse retinas (Figure 6—figure supplement 2A [rightmost column] and 2C). However, retinal rosettes and Tsc1-deficient retinal clonal expansion were still observed in about three-quarters of Psmb-tko;Tsc1fl/fl;Tyrp1-Cre mouse retinas (Figure 6—figure supplement 2A [second rightmost column] and 2C). These results suggest that there are additional regulator(s) of mTORC1-induced RPC clonal expansion, in addition to the immunoproteasomes.

In parallel to promoting protein degradation via the immunoproteasomes, mTORC1 also enhances the synthesis of many cellular proteins by activating S6 via S6K1 and by directly inactivating eukaryotic translation initiation factor 4E-binding protein 1 (Ben-Sahra and Manning, 2017; Saxton and Sabatini, 2017). Hif1a is among the proteins subjected to mTORC1-induced translational upregulation (Bernardi et al., 2006); it, in turn, induces the expression of various target genes to support mTORC1-induced cell proliferation, growth, and survival (Keith et al., 2011). We also found that the level of Hif1a, but not the level of its homolog, Hif2a, was significantly increased in Tsc1fl/fl;Tyrp1-Cre mouse retinas (Figure 6A). To investigate whether mTORC1 mediated Hif1a to generate the observed phenotypes, we deleted Hif1a concomitantly with Tsc1 in Tsc1fl/fl;Hif1afl/fl;Tyrp1-Cre mice. We found that 96% of Tsc1fl/fl;Hif1afl/fl;Tyrp1-Cre mouse retinas did not have rosette structures and MG cell loss (Figure 6B [third row] and C). R26tdTom-positive Cre-affected cell population was also decreased significantly in Tsc1fl/fl;Hif1afl/fl;Tyrp1-Cre mouse retinas compared to Tsc1fl/fl;Tyrp1-Cre littermate mouse retinas (Figure 6B [second row] and D), suggesting that the hyperexpansion of Tsc1-deficient cell clones was suppressed in these retinas. Furthermore, the telomere lengths of Tsc1fl/fl;Hif1afl/fl;Tyrp1-Cre mouse retinal cells were increased relative to those of Tsc1fl/fl;Tyrp1-Cre littermate mouse retinal cells (Figure 6E). The levels of the senescence markers induced in degenerating Tsc1fl/fl;Tyrp1-Cre mouse retinas were also decreased significantly in Tsc1fl/fl;Hif1afl/fl;Tyrp1-Cre mouse retinas (Figure 6A and B [two bottom rows]). These results suggest that Hif1a is necessary for the clonal hyperexpansion of Tsc1-deficient cells and the subsequent mitotic aging and senescence-associated degeneration of MG in Tsc1fl/fl;Tyrp1-Cre mouse retina.

Figure 6 with 2 supplements see all
Rescue of Tsc1fl/fl;Tyrp1-Cre mouse retinal phenotypes by concomitant deletion of hypoxia-induced factor 1-alpha (Hif1a).

(A) Levels of senescence markers in P30 mouse retinas with indicated genotypes were analyzed by Western blot (WB). Relative amounts of proteins used in each sample were determined by WB detection of β-actin. (B) Distributions of mTORC1-active cells, which are positive to pS6, MG, which are positive to Sox9, and senescent cells, which are positive to β-gal and p53, were examined by immunostaining of the eye sections. Expression of those markers in the Cre-affected cells were determined by comparing the expression of the Cre reporter R26tdTom. (C) Numbers of Sox9-positive MG in the mouse retinas were counted and the relative numbers are shown in the graph. (D) R26tdTom-negative wild-type cells and R26tdTom-positive Cre-affected cells in the retinas were determined byFACS and shown in the graph. (E) The lengths of telomeres of the cells isolated from P30 mouse retinas with the indicated genotypes were compared with that of control sample, and their relative values are shown in the graph. Error bars denote SD. Numbers of samples analyzed are shown in the graphs. n.s., not significant; *p < 0.05; **p < 0.01; ***p < 0.001.

Hif1a supports the accelerated cell cycle progression of Tsc1-deficient RPCs

We next investigated whether Hif1a supports the overproliferation of Tsc1-deficient RPCs and thus the hyperexpansion of Tsc1-deficient retinal clones. We measured the numbers of proliferating cells that incorporated a thymidine analog 5-ethynyl-2′-deoxyuridine (EdU) during S-phase of the cell cycle and then progressed to G2/M-phases by accumulating phosphorylated histone H3 (pH3) or incorporated two thymidine analogs, 5-chloro-2'-deoxyuridine (CldU) and 5-iodo-2'-deoxyuridine (IdU), serially in their first and second S-phases (Figure 7A). We found that the numbers of EdU;pH3-positive and CldU;IdU-positive RPCs were significantly elevated in P0 Tsc1fl/fl;Tyrp1-Cre mouse retinas in comparison to those in Tsc1fl/+;Tyrp1-Cre littermate mouse retinas (Figure 7A [third and seventh rows] and C). The numbers of EdU;Otx2-positive cells, which exited the cell cycle for differentiation to PRs in 12 hr, were also increased in the Tsc1fl/fl;Tyrp1-Cre mouse retinas (Figure 7A [fifth row] and C). Consequently, the sizes of R26tdTom-positive Tsc1-deficient cells in P0 Tsc1fl/fl;Tyrp1-Cre mouse retinas were significantly increased compared to those of R26tdTom-positive Tsc1-heterozygous retinal cells in Tsc1fl/+;Tyrp1-Cre littermates (Figure 7A [top row] and B). However, the numbers of proliferating cells and the sizes of R26tdTom-positive Tsc1-deficient clones were decreased significantly in P0 Tsc1fl/fl;Hif1afl/fl;Tyrp1-Cre littermate mouse retinas (Figure 7A [right most columns] - C). These results therefore suggest that Hif1a is necessary for the accelerated cell cycle progression of Tsc1-deficient mouse RPCs and consequent expansion of Tsc1-deficient retinal clones.

Figure 7 with 1 supplement see all
Hif1a supports mTORC1-induced retinal progenitor cell (RPC) proliferation through the expression of glycolytic enzymes.

(A) The Cre-affected cells in P0 mouse retinas with the indicated genotypes were visualized by R26tdTom Cre reporter. Distribution of mTORC1-active cells in the boxed areas was also examined by immunostaining of pS6 and shown in the second row. Proliferation and cell cycle progression of RPCs in the mouse retinas were also examined by 5-ethynyl-2′-deoxyuridine (EdU)/5-chloro-2'-deoxyuridine (CldU) labeling and chasing experiments as described in Materials and methods. (B) R26tdTom-positive Cre-affected cell population in the mouse retinas were quantified by FACS and shown in the graph. Error bars, SD. Numbers of samples analyzed are shown in the graph (four independent litters). (C) RPCs that had progressed from S-phase to G2/M-phase for 3 hr (EdU;pH3-positive); PRs that had been born for 12 hr (EdU;Otx2-positive); and RPCs that had reentered cell cycle for 15 hr (CldU;IdU-positive) in the mouse retinas were counted and shown in the graph. Error bars denote SD (n = 5, five independent litters). *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. (D) Relative mRNA levels of Hif1a, hexokinase 2 (Hk2), and pyruvate kinase M2 (Pkm2) in P0 mouse retinas with the indicated genotypes were examined by real-time quantitative polymerase chain reaction (RT-qPCR). The y-axis values are relative 2-ΔCt values against those of Tsc1fl/+;Tyrp1-Cre mouse retinas. Error bars, SD (n = 4, three independent litters). (E) Expressions of the indicated proteins in the mouse retinas were examined by Western blot (WB). Relative amounts of the proteins used in each sample were determined by WB detection of β-actin. (F) Relative WB band intensities of the proteins are determined by the ImageJ software and shown in the graph. Error bars, SD (n = 3, three independent litters). (G) Distributions of Pkm2, Tom20, and hypoxyprobe-labeled proteins in the mouse retinas were examined by immunostaining. (H) P0 Tsc1fl/fl;Tyrp1-Cre littermate mice were injected with a chemical inhibitor of glycolysis (2DG) or mitochondrial oxidative phosphorylation (Metformin), and the eye sections of the mice were obtained at 12 hr post-injection for the immunodetection of RPCs that had proliferated (CldU-positive) and/or reentered second cell cycle (CldU;IdU-positive) for 15 hr. Relative numbers of those cells in the mouse retinas were shown in the graphs in (I) and (J), respectively. Error bars, SD (n = 5, four independent litters). (K) Schematic diagram depicts the mTORC1-Hif1a-glycolysis cascade in developmental RPC clonal hyperexpansion, which leads to senescence-associated degeneration of MG in the mature retina.

Hypoxia and induction of Hif1a in Tsc1fl/fl;Tyrp1-Cre mouse retinas

We found that the level of Hif1a protein, but not that of Hif1a mRNA, was increased significantly in P0 Tsc1fl/fl;Tyrp1-Cre mouse retinas (Figure 7D and E). Hif1a is known to be accumulated under hypoxia by being resistant from proteasomal degradation (Ivan et al., 2001; Jaakkola et al., 2001). Therefore, Hif1a elevation could result from mTORC1-induced translational upregulation and/or hypoxia. We, thus, examined whether a hypoxic condition was established in the mouse retinas using the hypoxyprobe (Varghese et al., 1976), and found that P0 Tsc1fl/fl;Tyrp1-Cre mouse retinas were more hypoxic than Tsc1fl/+;Tyrp1-Cre littermate retinas (Figure 7E and G [bottom row images]). Interestingly, although hypoxia can compromise mitochondrial ATP synthesis (Vander Heiden et al., 2009), the level of ATP was significantly elevated in P0 Tsc1fl/fl;Tyrp1-Cre mouse retinas compared to that in Tsc1fl/+;Tyrp1-Cre littermate control retinas, and it returned to the control level in P0 Hif1afl/fl;Tsc1fl/fl;Tyrp1-Cre mouse retinas (Figure 7—figure supplement 1A). These results suggest that Tsc1-deficient RPCs mediate Hif1a to supply the ATP, which is necessary for their accelerated cell cycle progression and cell growth under the hypoxic condition of Tsc1fl/fl;Tyrp1-Cre mouse retinas.

Hif1a-induced glycolytic gene expression is necessary for the proliferation of Tsc1-deficient RPCs

It is known that glycolysis plays important roles in cellular ATP synthesis under hypoxia (Cerychova and Pavlinkova, 2018). Consistent with this, the level of lactate, which accumulates when glycolysis occurs in a condition of inefficient mitochondrial respiration (Cerychova and Pavlinkova, 2018), was significantly elevated in P0 Tsc1fl/fl;Tyrp1-Cre mouse retinas compared to Tsc1fl/+;Tyrp1-Cre littermate mouse retinas, but not in Hif1afl/fl;Tsc1fl/fl;Tyrp1-Cre littermate mouse retinas (Figure 7—figure supplement 1B). This suggests that Hif1a is necessary for the enhanced glycolysis observed in Tsc1fl/fl;Tyrp1-Cre mouse retinas.

We further examined whether Hif1a contributes to glycolytic ATP synthesis in Tsc1-deficient cells by inducing the expression of the glycolytic enzymes, such as hexokinase 2 (Hk2) and pyruvate kinase M2 (Pkm2), which are transcriptional targets of Hif1a (Düvel et al., 2010). We found that the mRNA and protein levels of those glycolytic enzymes were significantly elevated in P0 Tsc1fl/fl;Tyrp1-Cre mouse retinas compared to Tsc1fl/+;Tyrp1-Cre littermate retinas (Figure 7D-F). The enzyme levels in P0 Tsc1fl/fl;Hif1afl/fl;Tyrp1-Cre mouse retinas were decreased and became similar to those in P0 Tsc1fl/+;Tyrp1-Cre mouse retinas. Our findings suggest that Hif1a plays a critical role in elevating these glycolytic enzymes in Tsc1fl/fl;Tyrp1-Cre mouse retinas.

Next, to delineate the causal relationship between glycolysis and the hyperproliferation of Tsc1-deficient RPCs, we injected 2-deoxy-D-glucose (2DG), a chemical inhibitor of glycolysis (Wick et al., 1957), or metformin, a chemical inhibitor of the mitochondrial respiratory chain I (Owen et al., 2000) and/or glycerophosphate dehydrogenase in the mitochondrial electron transport chain (Madiraju et al., 2014), to P0 Tsc1fl/+;Tyrp1-Cre and Tsc1fl/fl;Tyrp1-Cre mice. We found that 2DG treatment decreased the lactate level in Tsc1fl/+;Tyrp1-Cre mouse retinas while metformin increased this parameter (Figure 7—figure supplement 1C). Moreover, the lactate level in 2DG-treated Tsc1fl/fl;Tyrp1-Cre became similar to that of PBS-treated Tsc1fl/+;Tyrp1-Cre mouse retinas, suggesting that the lactate accumulation in Tsc1fl/fl;Tyrp1-Cre was caused by enhanced glycolysis. We also found that not only total numbers of CldU-labeled proliferating RPCs but also those that had reentered the cell cycle (assessed by monitoring the incorporation of IdU, such that the cells were CldU;IdU-positive) were decreased significantly in 2DG-treated Tsc1fl/fl;Tyrp1-Cre mouse retinas in comparison to the numbers in the PBS- and metformin-injected groups and became similar numbers observed in Tsc1fl/+;Tyrp1-Cre littermate mouse retinas (Figure 7H-J). These results suggest that Tsc1-deficient RPCs depend on glycolysis for their accelerated cell cycle progression.

Collectively, our results suggest that mTORC1 increases the cellular Hif1a protein level through both translational and post-translational mechanisms in the RPCs of Tsc1fl/fl;Tyrp1-Cre mouse retinas by generating a hypoxic condition. This elevated Hif1a facilitates the expression of glycolytic enzymes to supply the ATP necessary for RPC proliferation (Figure 7K). This enables the RPCs to expand rapidly to exceed their mitotic division limit prior to producing the MG, which degenerate from senescence-associated cell death in mature Tsc1fl/fl;Tyrp1-Cre mouse retina.

Discussion

The growth and differentiation of vertebrate nervous tissues are tightly controlled in specific spatial and temporal manners to ensure the proper formation of their unique three-dimensional structures. During the growth of the double-layered optic cup of vertebrates, cell proliferation is significantly faster at the inner cup layer compared to the outer cup layer (Moon et al., 2018). Here, we detected mTORC1 activity only in the inner optic cup layer (Figure 1—figure supplement 1), suggesting that mTORC1 may involve in the spatially differentiated growth of the optic neuroepithelial continuum. In support of this, hamartomatous malformation of the CB and iris were reported in human TSC patients and in mice lacking Tsc1 throughout the optic cup (Eagle et al., 2000; Hägglund et al., 2017; Milea and Burillon, 1997). We also found CB/iris malformation in Tsc1fl/fl;Rax-Cre mice, in which Tsc1 was lost throughout the optic neuroepithelium, and in Tsc1fl/fl;Tyrp1-Cre mice, in which Tsc1 was lost in the RPE, CM, and peripheral retinal clones. In contrast, we did not observe CB/iris defects in mice lacking Tsc1 only in the retina (i.e., Tsc1fl/fl;Chx10-Cre mice) and RPE (i.e., Tsc1fl/fl;Mlana-Cre mice) (Figure 3A). This suggests that TSC plays an essential role in the CM to generate the CB and iris by suppressing the growth of the areas. However, the deletion of Tsc2 from the RPE and CM of Tsc2fl/fl;Tyrp1-Cre mice did not cause CB/iris malformations, although the mice shared the retinal phenotypes of Tsc1fl/fl;Tyrp1-Cre mice (Figure 1—figure supplement 4). These results suggest that Tsc1 regulates CB and iris development independent of the mTORC1 pathway. Given reported findings that Tsc1 functions in the tumor growth factor β (TGFβ)-Smad pathway (Thien et al., 2015) and the activation of this pathway in mouse CB (Srinivasan et al., 1998), we speculate that Tsc1 might contribute to CB and iris development via the TGFβ-Smad pathway. In parallel, we show Tsc1 also cooperates with Tsc2 to suppress mTORC1 in CM RPCs, whose cell division rate should be maintained at low to avoid of precocious arrival of the RPCs to the division limit. Collectively, these results suggest that Tsc1 is a multifunctional regulator of vertebrate eye development.

Our data suggest that the deletion of Tsc1 in a small RPC subset derived from the CM can cause the degenerative phenotypes in the mature retina (Figure 1). The CM RPCs in mouse retina were found to proliferate less robustly than the majority RPCs in the central retina (Belanger et al., 2017; Marcucci et al., 2016), so did the RPCs in lower vertebrate CMZ (Harris and Perron, 1998). However, the CM RPCs could have a competitive advantage for clonal expansion over the majority wild-type RPCs, which divide and expand slowly, when their cell cycle progression is accelerated by losing Tsc1 (Figure 4—figure supplement 4). This made the Tsc1-deficient RPCs divide exceeding the division limit and cause the degenerative phenotypes in Tsc1fl/fl;Tyrp1-Cre mouse retinas. The competitive hyperexpansion of a Tsc1-deficient RPC exceeding the division limit did not occur in Tsc1fl/fl;Rax-Cre and Tsc1fl/fl;Chx10-Cre mouse retinas, since its neighboring RPCs also lose Tsc1 and expand as fast as the RPC (Figures 4C and 5A; Figure 4—figure supplement 4). Therefore, Tsc1-deficient RPCs should be present in a minor population to have a competitive advantage in the clonal expansion.

The competitive clonal expansion could also occur in other conditions, where two RPC populations expand at different speeds. The Rptorfl/fl;Tyrp1-Cre mouse retina, which is composed of relatively fast dividing wild-type RPCs and slowly dividing (or cell cycle arrested) Rptor-deficient RPCs, could be an example. However, wild-type RPCs in the Rptorfl/fl;Tyrp1-Cre mouse retina did not overproliferate even though their neighboring Rptor-deficient RPCs stopped expanding their clones (Figure 5F–J). Therefore, the overproliferation might be a specific feature of Tsc1-deficient RPC but not a relative property obtained by the division differences.

The MG are among the last-born retinal cell types, arising around the first postnatal week in mice (Cepko, 2014; Hoang et al., 2020). Therefore, given the specific temporal sequence of retinal histogenesis, RPCs producing MG are likely to be mitotically older than those producing other retinal cell types. Since Tsc1-deficient RPCs overproliferate during the embryonic and early postnatal periods (Figure 4E), the mitotic ages of Tsc1-deficient RPCs in Tsc1fl/fl;Tyrp1-Cre mice should be far greater than those of wild-type RPCs when they are producing MG. Consequently, the MG born from Tsc1-deficient RPCs might be degenerated due to aging-related changes (Figure 5D and E). However, we found little MG degeneration or rosette formation in the retinas of old (2 years) wild-type C57BL/6J mice (data not shown), suggesting that the phenotypes of Tsc1fl/fl;Tyrp1-Cre and Tsc2fl/fl;Tyrp1-Cre mouse retinas are not simple aging-related changes.

Among the last-born retinal cell types, only MG exhibited significant degeneration in Tsc1fl/fl;Tyrp1-Cre mouse retina (Figure 1D). The MG might be in less terminal stage than the other last-born types (i.e., rPR and BC) in terms of differentiation, since they can resume cell cycle to regenerate neurons in the injured cold-blooded vertebrate retinas (Lahne et al., 2020). Mouse MG could also divide after the injury, if histone deacetylase inhibitor is provided after viral expression of a proneural transcription factor achaete-scute homolog 1 (Jorstad et al., 2017). Furthermore, given the fact that developing cells are more sensitive than terminally differentiated cells to cell death (Fuchs and Steller, 2011; Vaux and Korsmeyer, 1999), MG are likely more sensitive to the cell death than rPR and BC. However, the mechanisms underlying the differential apoptotic sensitivity of these last-born retinal cell types should be investigated in future studies.

The factors responsible for inducing the death of Tsc1-deficient RPCs and/or MG in Tsc1fl/fl;Tyrp1-Cre mouse retinas remain unclear. The repetitive division of RPCs might exploit telomeres beyond the critical limit and result in mitotic catastrophe, leading to senescence-associated cell death (Cleal and Baird, 2020). However, telomere shortening did not appear to be a direct cause of MG degeneration in the mouse retinas, since it was not suppressed by Tert overexpression (Figure 5—figure supplement 2). Furthermore, unlike the strong relationship between telomere shortening and senescence in human cells, telomere shortening in mouse cells is rarely associated with deleterious effects because the mouse telomere is 5–10 times longer than the human telomere (Calado and Dumitriu, 2013). Therefore, the senescence observed in Tsc1fl/fl;Tyrp1-Cre mouse retina might be caused by other cellular alterations associated with the RPC overproliferation.

The activation of mTORC1 by Tsc1 deletion was also shown to cause premature aging in mouse intestinal stem cells (ISCs) (He et al., 2020). The aging of Tsc1-deficient ISCs was reported to mediate the stress-induced p38 mitogen-activated protein kinase (MAPK) pathway, which activates p53 to block ISC proliferation. We also observed the activation (i.e., phosphorylation) of p38 MAPK in all four Tsc1-cko mouse retinas (data not shown), whereas senescence occurred only in Tsc1fl/fl;Tyrp1-Cre mouse retinas (Figure 5C). The activation of p53 did not likely cause degeneration of Tsc1-deficient RPCs and/or MG, either, since Trp53fl/fl;Tsc1fl/fl;Tyrp1-Cre mice still exhibited MG degeneration and retinal rosette formation (data not shown). Therefore, the activations of p38 MAPK and p53 may represent a cellular stress condition induced by mTORC1 hyperactivation, which increases cellular metabolism to elevate reactive oxygen species as byproducts (Ben-Sahra and Manning, 2017), but do not appear to trigger the senescence-associated cell death in Tsc1fl/fl;Tyrp1-Cre mouse retinas. A previous study further showed that intestinal epithelial cells (IECs) derived from Tsc1-deficient ISCs were degenerated via receptor interacting protein kinase 3 (Ripk3)-mediated necroptosis (Xie et al., 2020). However, in contrast to the critical roles of Ripk3 in the degeneration of Tsc1-deficient IECs, the MG degeneration in Tsc1fl/fl;Tyrp1-Cre mice was not blocked by concomitant deletion of Ripk3 (data not shown). These results suggest that the mechanism of MG degeneration in Tsc1fl/fl;Tyrp1-Cre mice might be different from the stress-induced aging and degeneration of IECs, and thus further investigation is warranted.

Embryonic and neonatal mouse retinas receive oxygen that diffuses from the distant hyaloid vessels in the vitreous and the choroid vessels across the RPE, whereas the mature mouse retina has close access to the oxygen released from intraretinal blood vessels (Lutty and McLeod, 2018; Saint-Geniez and D’amore, 2004). RPCs in the inner embryonic mouse retinas are farthest from the vessels, and are likely to exist under more hypoxic conditions than cells adjacent to the vitreous and RPE. The inner RPCs therefore might adapt to use glycolysis ATP synthesis than oxygen-dependent mitochondrial ATP synthesis for proper proliferation, as has been suggested for Xenopus retina (Agathocleous et al., 2012). Furthermore, Tsc1-deficient RPCs might spend more ATP for fast growth and division, compared to wild-type RPCs (Figure 7—figure supplement 1A), making their mitochondria utilize more oxygen to supply ATP. This is expected to worsen the hypoxia of Tsc1fl/fl;Tyrp1-Cre mouse RPCs and thereby compromise their mitochondrial ATP production eventually. In this situation, glycolytic ATP synthesis compensates for the ATP shortage in hypoxic Tsc1-deficient RPCs through the mTORC1-induced enhanced translation and stabilization of Hif1a, which, in turn, increases the expression of glycolytic enzymes (Figure 7D–F). The glycolytic gene expression is therefore decreased in the absence of Hif1a, such that ATP cannot be produced at a level sufficient to support mTORC1-induced cell proliferation and growth. This might decelerate the growth and proliferation of Tsc1-deficient RPCs and normalize the speed of retinal development.

Materials and methods

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
AntibodyAnti-Aquaporin(AQP1)(Mouse monoclonal)Novus BiologicalsNB600-749IHC (1:200)
AntibodyAnti-beta-actin(Rabbit polyclonal)Santa Cruz Biotechnologysc-1616WB (1:1000)
AntibodyAnti-beta-catenin(Rabbit polyclonal)Cell Signaling Technology9562IHC (1:200)
AntibodyAnti-beta-galactosidase(Mouse monoclonal)Developmental Studies Hybridoma Bank (DSHB)JIE7WB (1:500)IHC (1:100)
AntibodyAnti-BrdU(CldU)(Rat monoclonal)Novus BiologicalsNB500-169IHC (1:200)
AntibodyAnti-BrdU(IdU)(Mouse monoclonal)EXBIO11–286 C100IHC (1:200)
AntibodyAnti-Brn3b(Rabbit polyclonal)Santa Cruz Biotechnologysc-31989IHC (1:200)
AntibodyAnti-Calbindin(Rabbit polyclonal)Swant Inc.CB-38IHC (1:200)
AntibodyAnti-Cdo(Goat polyclonal)R&D SystemsAF2429IHC (1:200)
AntibodyAnti-c-myc(Mouse monoclonal)Santa Cruz Biotechnologysc40WB (1:1000)
AntibodyAnti-Cleaved caspase-3(Rabbit polyclonal)Cell Signaling Technology9661IHC (1:200)
AntibodyAnti-Ezrin(Mouse monoclonal)Invitrogen Biotechnology35–7300IHC (1:200)
AntibodyAnti-GFAP(Rabbit polyclonal)Abcamab48050IHC (1:200)
AntibodyAnti-GFP(Chicken polyclonal)Abcamab13970IHC (1:2000)
AntibodyAnti-Glutamine synthase (Rabbit polyclonal)Sigma-AldrichG-2781IHC (1:200)
AntibodyAnti-Hif1a(Mouse monoclonal)R&D SystemsMAB1536WB (1:1000)
AntibodyAnti-Hif2α/Epas1(Rabbit polyclonal)Novus BiologicalsNB100-122WB (1:1000)
AntibodyAnti-Hk2(Rabbit polyclonal)Cell Signaling Technology2867WB 1:1,000IHC (1:200)
AntibodyAnti-Hydroxy-Hif1alpha(Rabbit polyclonal)Cell Signaling Technology3434WB (1:1000)
AntibodyAnti-M-opsinMerck MilliporeAB5405IHC (1:200)
AntibodyAnti-Otx2(Rabbit polyclonal)Abcamab25985IHC (1:200)
AntibodyAnti-Otx2(Rabbit polyclonal)Abcamab183951IHC (1:200)
AntibodyAnti-Otx2(Goat polyclonal)R&D SystemsAF1979-SPIHC (1:200)
AntibodyAnti-p21/Cip1(Mouse monoclonal)Santa Cruz BiotechnologySC817WB (1:1000)IHC (1:200)
AntibodyAnti-p53(Mouse monoclonal)Merck MilliporeOP03-100WB (1:1000)IHC (1:200)
AntibodyAnti-Pax6(Rabbit polyclonal)CovancePRB-278PIHC (1:200)
AntibodyAnti-Phospho Smad1/5(S463/465)(Rabbit polyclonal)Invitrogen Biotechnology700047IHC (1:200)
AntibodyAnti-phospho Smad2(ser465/467)(Rabbit polyclonal) -Cell Signaling Technology18338IHC (1:200)
AntibodyAnti-phospho-Histone H3(S10) (pH3; Rabbit polyclonal)Merck Millipore04–1093IHC (1:200)
AntibodyAnti-phospho-S6(S235/236) (pS6; Rabbit polyclonal)Cell Signaling Technology2211WB (1:1000)IHC (1:200)
AntibodyAnti-Psmb8(Mouse monoclonal)Enzo Life ScienceBML-PW8845WB (1:1000)
AntibodyAnti-Psmb9(Mouse monoclonal)Santa Cruz Biotechnologysc-373996WB (1:1000)
AntibodyAnti-Psmb10(Rabbit polyclonal)Abcamab1183506WB (1:1000)
AntibodyAnti-Raptor(Rabbit polyclonal)Cell Signaling Technology2280WB (1:1000)
AntibodyAnti-PKM2(Rabbit polyclonal)Abgentap7173dWB (1:1000)
AntibodyAnti-PKM2(Rabbit polyclonal)Cell Signaling Technology4053IHC (1:200)
AntibodyAnti-Recoverin(Rabbit polyclonal)ChemiconAB5585WB (1:1000)
AntibodyAnti-RPE65(Mouse monoclonal)Abcamab13826WB (1:1000)
AntibodyAnti-Rhodopsin(Mouse monoclonal)ChemiconMAB5356IHC (1:200)
AntibodyAnti-S6(Mouse monoclonal)Cell Signaling Technology2317WB (1:1000)
AntibodyAnti-Tom20(Rabbit polyclonal)Santa Cruz Biotechnologysc-11415WB (1:1000)
AntibodyAnti-Tsc1(Rabbit polyclonal)Cell Signaling Technology4906WB (1:1000)
AntibodyAnti-Tsc2(Rabbit polyclonal)Cell Signaling Technology3612WB (1:1000)
AntibodyAnti-Tubulin-ßIII (Tuj1; Mouse monoclonal)CovanceMMS-435PIHC (1:200)
AntibodyAnti-Vsx2(Mouse monoclonal)Santa Cruz Biotechnologysc365519WB (1:1000)IHC (1:200)
Genetic reagent (Mus musculus)B6.129-Hif1atm3Rsjo/JJackson Laboratory007561Hif1аfl/fl
Genetic reagent (Mus musculus)B6.Cg-Rptortm1.1Dmsa/JJackson Laboratory013188Rptorfl/fl
Genetic reagent (Mus musculus)Tsc1tm1Djk/JJackson Laboratory005680Tsc1fl/fl
Genetic reagent (Mus musculus)Tsc2tm2.1Djk/MmjaxJackson Laboratory37154-JAXTsc2fl/fl
Genetic reagent (Mus musculus)CAG-loxP-3xpolyA-loxP-mTERT-IRES-Hygro-rHidema et al., 2016dTert
Genetic reagent (Mus musculus)Tg(Tyrp1-cre)1IpcMori et al., 2002Tyrp1-Cre
Genetic reagent (Mus musculus)Tg(Mlana-cre)5BeeAydin and Beermann, 2011Mlana-Cre
Genetic reagent (Mus musculus)Tg(Chx10-EGFP/cre,-ALPP)2ClcRowan and Cepko, 2004Chx10-Cre
Genetic reagent (Mus musculus)Tg(Pax6-cre,GFP)2PgrMarquardt et al., 2001Pax6-аCre
Genetic reagent (Mus musculus)Tg(Rax-cre)1ZkozKlimova et al., 2013Rax-Cre
Genetic reagent (Mus musculus)Tg(Slc1a3-cre/ERT)1Nat/JJackson Laboratory012586Slc1a3-CreERT2
Genetic reagent (Mus musculus)B6.129 × 1-Gt(ROSA)26Sortm1(EYFP)Cos/JJackson Laboratory006148R26EYFP
Genetic reagent (Mus musculus)B6.Cg-Gt(ROSA)26Sortm14(CAG-tdTomato)Hze/JJackson Laboratory007914R26tdTomato
Genetic reagent (Mus musculus)Tg(Tyrp1-cre/ERT2)1JwkThis paperTyrp1-CreERT2
Genetic reagent (Mus musculus)B6-Psmb8em1hwl Psmb9em1hwl/KorlThis paperPsmb8-/-,9-/-
Genetic reagent (Mus musculus)B6-Psmb10em1JwkThis paperPsmb10-/-
Chemical compound, drugPhalloidin-AlexaFluor 647Abcamab1767591:200
Chemical compound, drugHoechst 33,342InvitrogenH13991:1,000
Chemical compound, drugClick-iT EdU Cell Proliferation Kit for Imaging, Alexa Fluor 647 dyeInvitrogenC10340
Chemical compound, drug2-Deoxy-D-glucoseSigma-AldrichD6134
Chemical compound, drugMetformin hydrochlorideSigma-AldrichPHR1084
Chemical compound, drug2-Chloro-5-deoxyuridine (CldU)Sigma-AldrichC6891
Chemical compound, drug5-Iodo-2′-deoxyuridineSigma-AldrichI7125
Chemical compound, drugMethyl celluloseSigma-AldrichM7140
Chemical compound, drugL-GlutamineGIBCO25030–081
Chemical compound, drugB27 supplementGIBCO17504–044
Chemical compound, drugN2 supplementGIBCO17502–048
Chemical compound, drugHeparinMilliporeM535142
Chemical compound, drugProtease inhibitorMilliporeM535142
Peptide, recombinant proteinDnase ISigma-AldrichDN25-100mg
Peptide, recombinant proteinEpidermal growth factor (EGF)Sigma-AldrichE4127
Peptide, recombinant proteinFibroblast growth factor 2(hBFGF)Sigma-AldrichF0291
Commercial assay or kitENLITEN ATP Assay SystemPromegaFF2000
Commercial assay or kitLactate-Glo AssayPromegaJ5021
Commercial assay or kitSuperSignal West Pico plus chemiluminescent substrateThermo Scientific34580
Commercial assay or kitSuperSignal West Femto plus chemiluminescent substrateThermo Scientific34095
Commercial assay or kitIn situ Cell death Detection Kit, TMR-redRoche12156792910
Commercial assay or kitIn situ Cell Death Detection Kit, Fluorescein (TUNEL green)Roche11684 795910
Commercial assay or kitHypoxyprobe KitHypoxyprobeHP1-100kit
Software, algorithmFluoview 4.0Olympus CorporationN/A
Software, algorithmImaris 9.3BitplaneN/A
Software, algorithmGraphPad Prism v7.0GraphPad softwareN/A
Sequence-based reagentActinß1-forwardBioneerRT-qPCR primerCTGGCTCCTAGCACCATGAAGAT
Sequence-based reagentActinß1-reverseBioneerRT-qPCR primerGGTGGACAGTGAGGCCAGGAT
Sequence-based reagentPkm2-forwardBioneerRT-qPCR primerTCGCATGCAGCACCTGATT
Sequence-based reagentPkm2-reverseBioneerRT-qPCR primerCCTCGAATAGCTGCAAGTGGTA
Sequence-based reagentHk2-forwardBioneerRT-qPCR primerTGATCGCCTGCTTATTCACGG
Sequence-based reagentHk2-reverseBioneerRT-qPCR primerAACCGCCTAGAAATCTCCAGA
Sequence-based reagentHif1a-forwardBioneerRT-qPCR primerACCTTCATCGGAAACTCCAAAG
Sequence-based reagentHif1a-reverseBioneerRT-qPCR primerCTGTTAGGCTGGGAAAAGTTAGG
Sequence-based reagentPsmb8-forwardBioneerGenotyping primerTTGGTACTGTGGCTTTCGCTTTC
Sequence-based reagentPsmb8-reverseBioneerGenotyping primerACACTCCTTCCTCTGTGCCACC
Sequence-based reagentPsmb9-forwardBioneerGenotyping primerGACCTTGAGTCGGTCACCTCC
Sequence-based reagentPsmb9-reverseBioneerGenotyping primerCACTTAGGGCCACCAGCTTCC
Sequence-based reagentPsmb10-forwardBioneerGenotyping primerACGCGAGTCACCCCA ATGTTT
Sequence-based reagentPsmb10-reverseBioneerGenotyping primerCGCCACAACCGAATCGTTAGT
Sequence-based reagentPsmb8-gRNA forwardBioneerCRISPR/Cas9TCGGGGGCAGCGGCCCGAGTGGG
Sequence-based reagentPsmb8-gRNA reverseBioneerCRISPR/Cas9CCAGGGCAGCCCACTCGGGCCGC
Sequence-based reagentPsmb9-gRNA forwardBioneerCRISPR/Cas9GGAGTTTGACGGGGGTGTCGTGG
Sequence-based reagentPsmb9-gRNA reverseBioneerCRISPR/Cas9CCCACCACGACACCCCCGTCAAA
Sequence-based reagentPsmb10-gRNA #3 forwardBioneerCRISPR/Cas9CACCGAACACGTCCTTCCGGGACTT
Sequence-based reagentPsmb10-gRNA #3 reverseBioneerCRISPR/Cas9AAACAAGTCCCGGAAGGACGTGTTC
Sequence-based reagentPsmb10-gRNA #5 forwardBioneerCRISPR/Cas9CACCGCTGCCAGAGGAATGCGTCCT
Sequence-based reagentPsmb10-gRNA #5 reverseBioneerCRISPR/Cas9AAACAGGACGCATTCCTCTGGCAGC

Mice

Information of mouse strains used in the experiments is listed in Key resources table. Psmb8-/-,9-/-,10-/- triple knock-out (Psmb-tko) and Tyrp1-CreERT2 mice were generated in this study. The Psmb-tko mice were generated using the CRISPR/Cas9 system, as it was described in a previous report (Kim et al., 2017). The genetic manipulations resulted in 25, 8, and 22 nucleotide (nt) deletions in mouse Psmb8, Psmb9, and Psmb10 genes, respectively (Figure 6—figure supplement 2). The sequences of guide RNAs used for introducing the deletions are provided in Key resources table. The Tyrp1-CreERT2 mice were generated following the same procedure reported previously (Mori et al., 2002). The transgenic mice were established by microinjection of the 7.3 kb NotI DNA segment of Tyrp1-CreERT2 into C57BL/6 J mouse embryos (two-cell stage). These injected embryonic cells were then injected into the inner cell mass of ICR mouse embryos. The tails of mice born from the surrogate mice were isolated for the genotyping with the primers listed in Key resources table. Two resulting F1 chimeric male mice carrying the transgene were crossed to C57BL6/J female mice to obtain an F2 generation with the heterozygous Tyrp1-CreERT2 mice.

To generate the cko mice, the mice having the floxed (fl) target gene alleles were bred with the mice expressing various Cre recombinases. To tracing of the cells experienced the Cre-dependent recombination at target gene loci, the mice were crossed with the R26EYFP and R26tdTom mice, and the cells experienced the Cre-dependent recombination of target sites were visualized by the fluorescence of the reporters. Experiments using the mice were carried out according to the guidance of Institutional Animal Care and Use Committee (IACUC) of KAIST (KA-2014–20). All mice used in this study were maintained in a specific pathogen-free facility of KAIST Laboratory Animal Resource Center.

Cryosections and immunohistochemistry

Request a detailed protocol

Pregnant and postnatal mice were euthanized after the anesthesia by intraperitoneal injection of tribromoethanol (Avertin, Sigma). The euthanized postnatal mice were perfused with phosphate buffered saline (PBS, pH7.5) containing 0.1% heparin (Millipore) and then with 4% paraformaldehyde (PFA, Sigma) in PBS. The eyes were isolated from the mice for further fixation in 4% PFA/PBS solution at 4°C for 2 hr. The embryos were isolated from the pregnant mice and fixed in 4% PFA/PBS solution at 4°C for 4 hr. The eyes and embryonic heads were then transferred to 20% sucrose/PBS solution for the incubation at 4°C for 16 hr before embedding in the Tissue-Tek OCT compound (Sakura). Cryosections (10–14 μm) of the frozen embryonic heads and postnatal mouse retinas were obtained and stained with hematoxylin and eosin (H&E) solutions for histologic examinations.

For immunohistochemistry, the sections were incubated for 1 hr in a blocking solution containing 10% normal donkey serum in PBS containing 0.2% Triton X-100. The sections were incubated with the primary antibodies at 4°C for 16 hr, and further stained with Alexa488-, Cy3-, or Alexa647-conjugated secondary antibodies (Jackson ImmunoResearch Laboratories) at room temperature (RT) for 1 hr. The antibodies are listed in Key resources table. Fluorescent images were obtained by confocal microscope (Fluoview FV1000 and FV3000; Olympus).

For the detection of proliferating cells in the embryos or neonatal mouse retinas, pregnant and neonatal mice were injected with 5-bromo-2′-deoxyuridine (EdU; 50 mg/kg, ThermoFisher). EdU was detected by incubating in Click-iT EdU Alexa Fluor 647 (Thermo Fisher Scientific) according to manufacturer’s instructions. Alternatively, the mice were injected with 5-chloro-2′-deoxyuridine (CldU; 30 mg/kg, Sigma) and 5-iodo-2′-deoxyuridine (IdU; 30 mg/kg, Sigma) into their peritoneal cavity for the indicated time periods prior to sample collection. The eyes were isolated from the mice and the eye sections were post-fixed in 4% PFA/PBS for 5 min, and washed three times with PBS with Triton X-100 0.2%. The sections were then incubated with 2 N HCl for 30 min, neutralized with borate buffer (pH 8.0) for 5 min (three times, 10  min) at room, followed by rinses with PBS (three times, 10  min). The sections were then subjected to the immunohistochemistry procedures described above.

Reverse transcription-quantitative polymerase chain reaction analysis

Request a detailed protocol

Total RNA from mouse retinas were isolated using easy-BLUE Total RNA Extraction Kit (iNtRON). About 1 μg of total RNA was reverse transcribed using SuperiorScript III Master Mix (Enzynomics) according to the manufacturer’s protocol. Real-time quantitative polymerase chain reaction (PCR) (RT-qPCR) was performed with TOPreal qPCR 2 X PreMIX (Enzynomics). The RT-qPCR was performed at 95°C for 10 min, followed by 50 cycles of 95°C for 15 s and 60°C for 15 s. Relative expression levels of gene of interest were calculated according to the 2−ΔΔCt method against β-actin mRNA. Primer sequences are listed in Key resources table.

Quantification of relative telomere lengths

Request a detailed protocol

Genomic DNA (gDNA) were extracted from mouse retinas by lysing the cells in DirectPCR Lysis Reagent (Viagen Biotech) supplemented with RNaseA (2 mg/ml, Bioneer) and proteinase-K (2 mg/ml, Roche), and then were purified further by the extraction with phenol-chloroform-isoamyl alcohol (25:24:1 (v/v); Sigma) followed by precipitation with 0.3 M (final) sodium acetate (pH 5.2) and 0.7 volume isopropanol. The lengths of telomeres in the gDNA dissolved in Tris-EDTA (pH 8.0) solution were then assessed by Relative mouse Telomere Length quantification qPCR assay kit (ScienCell Research Laboratories) according to the manufacturer’s protocol.

Western blot analysis

Request a detailed protocol

Total protein was extracted from mouse retinas with RIPA buffer (50 mM Tris [pH 8.0], 150 mM NaCl, 1.0% NP40, 0.5% sodium deoxycholate, 0.1% sodium dodecyl sulfate [SDS]) supplemented with the protease and phosphatase inhibitor cocktail (Roche). For separation of proteins, the cell lysates containing 30 μg proteins were analyzed by 8–15% SDS-polyacrylamide gel electrophoresis (SDS-PAGE) prior to the transfer onto polyvinylidene difluoride membranes. The membranes were blocked with 5% skim milk in Tris-buffered saline containing 0.1% Tween-20 (TBST, pH 7.4) at RT for 1 hr. The membranes were incubated at 4°C for 16 hr with primary antibodies listed in Key resources table, washed with the TBST at RT three times (10  min each), and incubated with the horseradish peroxidase-conjugated secondary antibodies at RT for 1 hr. Reactive bands were detected using SuperSignal West Pico Chemiluminescent Substrate or SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Fisher Scientific).

FACS analysis

Request a detailed protocol

Mouse retinas were collected in 1 ml Hank’s Balanced Salt Solution (HBSS; Life Technologies) after enucleating the lens from the eyes, and transferred into HBSS containing 0.1% Trypsin and DNase I (100  μg/ml; Sigma) followed by the incubation at 37°C for 30 min. Dissociated retinal cells were gently triturated in HBSS with 2% FCS, filtered through a 70 μm strainer before FACS analysis. The R26tdTom-positive cells were then analyzed by the BD Fortessa analyzer (BD Biosciences) or collected by the BD FACSAria cell sorter (BD Biosciences).

RPC neurosphere culture

Request a detailed protocol

The retinas of E13 mouse embryos were isolated and incubated in a dissociation solution (DMEM containing 0.1% Trypsin and DNase I [100 μg/ml; Sigma]) at 37°C for 30  min. The aggregates were further dissociated into single cells using Accutase and passed through a 70 μm cell strainer. The R26tdTom-positive cells were then collected by FACS for subsequent culture in DMEM/F12 medium containing 0.9% (w/v) methylcellulose matrix (Sigma), N2 supplement (GIBCO), B27(GIBCO), 2 mM L-glutamine (GIBCO), 100 U/ml penicillin, and 100 μg/ml streptomycin (GIBCO) supplemented with 10 ng/ml fibroblast growth factor 2 (Sigma) and 20 ng/ml epidermal growth factor (Sigma). Then the cells were cultured for 1–4 weeks and the number of primary neurospheres was counted every 3 days with microscopic observation.

ATP assay

Request a detailed protocol

ATP concentration was measured using the ENLIGHTEN ATP assay system (Promega). In brief, mouse retinas were homogenized in a buffer (0.25 M sucrose and 10 mM HEPES-NaOH, pH 7.4) and centrifuged at 3000 rpm at 4°C for 10 min. The supernatant (200 μl) was added to an equal volume of 10% trichloroacetic acid (Sigma) and centrifuged at 10,000 rpm at 4°C for 10 min. After centrifugation, 300 μl of the supernatant was added with 200 μl of neutralization buffer (1 M Tris-acetate buffer, pH 7.5) and then was diluted 30-folds in deionized water prior to the measurement of luminescence using the MICROLUMAT LB96P Reader (Berthold).

Lactate assay

Request a detailed protocol

Lactate concentration was measured using the Lactate-Glo Assay Kit (Promega) according to the manufacturer’s protocol. Briefly, 1 mg of the retina were homogenized in 400 μl of homogenization buffer (50 mM Tris, pH 7.5) with 50 μl of inactivation solution (0.6 N HCL) and immediately added to 50 μl of neutralization buffer. Before recording luminescence, 50 μl of samples were transferred into a white 96-well plate and added 50 μl of lactate detection reagent to the well, mix for 60 s, and incubate at RT for 1 hr. Luminescence was recorded with the MICROLUMAT LB96P Reader (Berthold).

Statistical analysis

Request a detailed protocol

Statistical analysis was performed by Prism Software (GraphPad) measurement tools. Data from statistical analysis are presented as the mean ± SD. Student’s t test was used to determine the significant difference between two genotypes and one-way ANOVA with Tukey’s post-test used to determine the significant differences among multiple groups. p-Values were calculated using a two-tailed unpaired t-test. p < 0.05 was considered statistically significant. *p < 0.05; **p < 0.01; ***p < 0.005; ****p < 0.001.

Data availability

All data generated or analysed during this study are included in the manuscript and supporting file; Source Data files have been provided for Figures 1 and 2.

References

    1. Harris WA
    2. Perron M
    (1998)
    Molecular recapitulation: the growth of the vertebrate retina
    Tl J Dev Biol 42:299–304.
    1. Mori M
    2. Metzger D
    3. Garnier JM
    4. Chambon P
    5. Mark M
    (2002)
    Site-specific somatic mutagenesis in the retinal pigment epithelium
    Vestigative Ophthalmology & Visual Science 43:1384–1388.
    1. Varghese AJ
    2. Gulyas S
    3. Mohindra JK
    (1976)
    Hypoxia-dependent reduction of 1-(2-nitro-1-imidazolyl)-3-methoxy-2-propanol by Chinese hamster ovary cells and KHT tumor cells in vitro and in vivo
    Cancer Research 36:3761–3765.
    1. Wick AN
    2. Drury DR
    3. Nakada HI
    4. Wolfe JB
    (1957)
    Localization of the primary metabolic block produced by 2-deoxyglucose
    The Journal of Biological Chemistry 224:963–969.
    1. Willbold E
    2. Rothermel A
    3. Tomlinson S
    4. Layer PG
    (2000)
    Müller glia cells reorganize reaggregating chicken retinal cells into correctly laminated in vitro retinae
    Glia 29:45–57.

Decision letter

  1. Paola Bovolenta
    Reviewing Editor; CSIC-UAM, Spain
  2. Carlos Isales
    Senior Editor; Medical College of Georgia at Augusta University, United States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Thank you for submitting your article "mTORC1-induced retinal progenitor cell over proliferation leads to accelerated mitotic aging and degeneration of descendent Müller glia" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Carlos Isales as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

There is general agreement that the authors report a fascinating phenotype. However, there are also several concerns that are listed in the individual reviewers' reports. The common and most pressing unresolved issues that the authors should experimentally address are the following:

1) Is the Trp1-Cre line somehow contributing to the phenotype? This could be solved by the use of theTrp1-CreERT2 line or the injection so AAV-Cre. Extensive discussion of this issue is provided in the individual reviewers' assessment of the manuscript

2) Can the results be interpreted in terms of cell competition? The study does not provide sufficient support to the statement that telomere shortening is a "cell autonomous" effect.

Reviewer #1:

This paper exposes the effects of inactivating mTOR in the developing retina. The authors make the surprising discovery that multiple deficits are observed when mTOR is deleted in a fraction of retinal cells, but not when it is knocked out in most/all retinal cells. They tie these context specific deficits to aberrant Hif1alpha activity and its effects on metabolism.

Strengths:

Fascinating phenotype, uniquely observed in the context of mixed Tsc1 knockout (KO) and wild type (WT) cells. The authors performed extensive genetics to demonstrate the unique link to this context

Extensive characterization demonstrates telemore shortening, mitotic exhaustion and senescence associated with the phenotypic defects.

Additional genetics, biochemistry and phenotyping show, convincingly, that the effects depend in part on the proteasome and, more so, on the ability of Hif1alpha to engage glycolytic metabolism.

Weaknesses:

The authors did not consider the possibility that the phenotype reflects cell competition between fitter mutant and weaker normal cells i.e. that the latter are critical for the phenotype.

Comments on the author's conclusions:

The authors show convincingly that the effects are due to mitotic exhaustion and senescence, and that they are dependent on Hif1alpha. The conclusion that the effect is "cell autonomous" seems unjustified, as the genetic data suggests that normal cells may be essential for effects in the mutant population.

Potential Impact:

The paper will be of considerably interest, not only to those in the field of retinal development, but to those studying the link between mTOR, metabolism and tissue growth and (despite it not being raised in the current version) to those studying cell competition.

The authors show that deleting Tsc1 in the ciliary margin(CM)/RPE with TRP1-Cre stimulates expansion of these cells, and secondary retinal defects, such as rosette formation. Interestingly, deleting Tsc1 with multiple other retinal Cre strategies did not have that effect (GLAST-Cre/Muller; Chx10-Cre/most retinal progenitors; MART1-Cre/RPE, a-Cre/peripheral retina, mRx-Cre/retina+CM), and even combining Chx10-Cre + MART1-Cre did not mimic the effect seen with TRP1-Cre. These data suggest the cause is linked to the expansion of minority Tsc1 null progenitors, and not the absence of Tsc1 in retinal cell types per se, which is a unique and fascinating, while also puzzling, finding. Quantification comparing het and homozygous mice revealed that, indeed, the biggest increase in KO cells is seen in with the TRP1-Cre mice, with some increase in α-Cre, indicating that it is likely that KO cells are out-packing wild type cells in models where there are a mixture of mutant and wt genotype progenitors. Cultured neurosphere studies showed that KO cells generate larger spheres than wt cells, but degenerate earlier. Studies with P30 retina showed that TRP-Cre KO, but not other strains, exhibit telomere shortening and senescence. In the contrary experiment, inhibiting mTOR by knocking out floxed Raptor lengthened telomeres specifically in the TRP1-Cre context. mTOR activates the proteasome, and deleting proteasome subunits partially rescued retinal defects. mTOR can induce Hif1alpha, and indeed this protein was induced as expected, and KO rescued cellular and telomere defects in the mTRP1-Cre Tsc1 KO retina. Additional studies confirm accelerated cell cycle progression, ATP production, glycolysis, induction of glycolytic enzyme mRNA/protein and dependency of both on Hif1alpha.

– It is interesting, but overlooked in the paper, that the telomere length in KO cells is not intrinsically shorter in KO cells (e.g. not in mRx-Cre KO or CHx10-Cre KO retina), but only in retinas with mixed wt and KO cells (significant in TRP1-Cre KO, and n.s. in α-Cre but mean is lower). That finding suggests that the telomere shortening may be dependent on the presence of neighboring WT cells. This could, for example, reflect "cell competition" between KO and WT cells, in which the elevated rates of division is due to greater fitness of KO cells, that is stimulated by their weaker neighbors. Indeed, mTOR signaling has been linked to cell competition in mammals: PMID: 29720666; PMID: 33652024. It should be feasible to assess this notion by isolating early stage KO cells and assessing whether their division rate and/or telomere attrition is boosted specifically in the presence of WT cells relative to pure KO cultures.

– After showing telomere shortening in the KO but not wt cells from the same TRP1-Cre retina, the papers states that this phenomenon must be "cell autonomous". If that is the case, why is there no telomere shortening in other strains in which the entire/most of the retia is KO? As noted above, it appears that the shortening may depend on WT neighbors, in which case the term "cell autonomous" would be inappropriate.

– The same is true of other biochemical phenotypes that are unique to the TRP1-Cre retina, notably senescence. It's feasible, therefore, that the entire basis for the phenotypic specificity in the TRP1-Cre retina is dependent on mutant/normal cell interaction. The authors should attempt assays to validate or rule out that possibility. In addition to the cell mixing studies above, it should also be feasible to perform clonal studies in explants or in vivo using retroviral Cre delivery. It would, for example, be feasible to at least assess whether the resulting clones contain senescent cells (e.g. SA-betagal-positive staining).

– The authors show that expressing Tert does not rescue elevated cell death in the TRP1-Cre KO retina. They conclude on page 12: "This suggests that telomere shortening is unlikely to be the cause of MG degeneration in Tsc1fl/fl;TRP1-Cre mice". They would need to demonstrate that telomere length is rescued to make that claim, otherwise they should modify to state: "This may mean either that Tert expression is insufficient to rescue telomere length in this context, or that telomere shortening does not cause MG degeneration"

Overall, the paper is really interesting, but the authors have overlooked an explanation for their results that should be addressed, given the prominent literature on cell competition, and its link to mTOR.

Reviewer #2:

The manuscript titled, "mTORC1-induced retinal progenitor cell overproliferation leads to accelerated mitotic aging and degeneration of descendent Müller glia" by Lim et al., reports several conditional knockouts of Tsc1 within a variety of ocular compartments. The authors conclude that loss of Tsc1 within a subset of retinal progenitor cells (RPCs) leads to overactivation of mTORC1 and a consequential acceleration of cell cycle progression. They further conclude that this enhanced proliferation leads to an exhaustion of the mitotic capacity of these RPCs that go on to differentiate into Müller glia (MG) that subsequently degenerate due to premature "aging". Finally, they implicate Hif1a-mediated glycolytic gene expression as being required for this phenotype. The authors employ an impressive battery of Cre lines and floxed alleles aimed at precisely defining the cellular origin of the Trp1-Cre dependent Tsc1 CKO phenotype as well as its molecular mechanism. The majority of the data are sound and reflect strict adherence to scientific rigor.

From the data, they make the following main conclusions:

1. Trp1-Cre+/tg; Tsc1flox/flox mice suffer from deregulated mTORC1 activity in a subset of peripheral, ciliary margin-derived RPCs and this results in "clonal expansion" of these mutant cells into the central wild-type retina. This phenotype is suggested to cause retinal laminar disorganization in the form of rosette structures.

2. Upon terminal differentiation of these rapidly diving CKO cells, MG are produced in excess, but eventually undergo apoptosis.

3. MG apoptosis is due to "mitotic aging" and resulting senescence-associated cell death.

4. These phenotypes are specific to the RPCs of the Trp1-Cre line and are not produced by other RPC Cre lines (Chx10-Cre, aPax6-Cre, and mRx-Cre).

5. Tsc1 CKO RPC hyperproliferation is dependent on HIF1a activity, which activates glycolysis required to drive that cellular state.

While authors present a large body of work, and they are clearly focused on identifying the precise molecular mechanism at play, it is still not clear to this reviewer what are primary versus secondary aspects of this phenotype. Lingering questions regarding the specificity of Trp1-Cre leave some doubt as to the accuracy of the interpretations provided in what is otherwise an experimentally sound paper.

The main, unresolved conundrum is the lack of RPC phenotype when using the Chx10-Cre, mRx-Cre, and aPax6-Cre lines. I understand the authors' argument that since Trp1-Cre is active in fewer RPCs, the CKOs cells may be allowed to expand into the territory of the more slowly dividing, centrally located WT cells. However, if this is indeed the case, why was it not observed in the Chx10-Cre or aPax6-Cre mice? My lab and others have used these line for years and we can state with absolute certainty that the Cres are mosaically expressed. Therefore, one would definitely end up with patches of CKOs cells adjacent to WT cells. This is particularly true for aPax6-Cre that is predominantly expressed in peripheral RPCs. Wouldn't the CKO patches overtake the WT patches centrally? The most parsimonious explanation for the discrepancies between the Trp1-Cre and the other RPC Cre lines, is that something other than Tsc1 CKO is contributing to the phenotype. I have several concerns regarding the Trp1-Cre line. Thanos, et al. (2012) reports that the line exhibits Cre toxicity leading to "RPE dysfunction and concomitant disorganization of RPE layer morphology, large areas of RPE atrophy, retinal photoreceptor dysfunction, and microglial cell activation in the affected areas". While the authors seem to be addressing this issue by comparing Trp1-Cre; Tsc1+/flox mice to Trp1-Cre; Tsc1flox/flox mice, are they certain that the mice in question all carry one copy of the Trp1-Cre transgene? This line is a random transgene insertion. Therefore, if one does not know where the transgene inserted into the genome, it's impossible to genotype for 1 versus 2 copies and a test cross would have to be used to determine zygosity. Different dosages of Cre could impact toxicity effects. Also, if the transgene inserted into an essential gene that could also lead to a phenotype in a homozygous state. The authors should also consider the genetic background of these mice as several rd alleles are known to be present within certain inbred strains (i.e. rd8 in C57BL/6N).

This reviewer recognizes the tremendous effort put forth in this paper and sincerely appreciates it. The utilization of mouse genetics is truly impressive, and I commend you for that. Nevertheless, I still cannot shake the concern that there is something about the Trp1-Cre line that is misleading you. It's very odd to me that the other RPC Cre lines show absolutely no phenotype in vivo. I understand that this is not an easy question to answer, but it would benefit your paper greatly if you could show some direct evidence that the Trp1-Cre mediated phenotype is specific to Tcs1 loss in RPCs without any contribution from negative effects of the Trp1-Cre itself. It's formally possible that Trp1-Cre on its own has an effect that's synergizing with Tsc1 loss to give you the phenotype.

You could employ the Trp1-CreERT2 line (Mori, et al., 2012). This an independent transgene and the random integration site is expected to be completely different from the Trp1-Cre transgene. Also, because Cre activity is tamoxifen-dependent, this line would be an effective means to confirm the Trp1-Cre Tsc1 CKO phenotype as being specific to Trp1-Cre mediated recombination of the floxed allele. In other words, the comparison would be Trp1-CreERT2; Tsc1flox/flox littermates with or without tamoxifen. However, there could still be issues with Cre toxicity upon induction. Another method could be AAV-Cre which could be used to sparsely infect RPCs with Cre followed by assessment of CKO cell expansion into WT areas. This is probably the preferred method as it does not depend on the Trp1 promoter.

Figure S1.

Here the authors show pS6 immunofluorescence to indicate expression in the retina but not the ciliary margin or RPE. Why is the retina staining so sparse? Wouldn't one expect more widespread expression with the proliferative RPC population? What is the identity of the pS6+ cells?

Figure 1.

Is the ciliary body actually "absent" as the authors state or is it malformed?

At P14 and P30, if MGs are missing due to apoptosis, why are other retinal cell types that require MGs for homeostatic support also not degenerative?

Might the MG loss actually just be a downregulation of Sox9 and GS expression during gliosis? GFAP, Cyclin D1, and p27 immunofluorescence could address this. However, it's also curious that in Figure S2 the authors don't show an increase in GFAP expression in the MG of the CKO mice. Such disruptions in retinal architecture would certainly be expected to result in a gliotic response of whatever MGs are still present near the affected area.

Figure 3.

The Sox9+ MG layer in the Chx10-Cre; Tsc1flox/flox mice looks very disorganized. Is this a representative image? If it is, it might reflect a subtle phenotype. The MG layer sometimes appears less organized when undergoing a gliotic response.

Figure 4.

Please indicate statistical significance in panels E, G, and H.

I take issue with the term "clonal expansion" as that's not really what this experiment examines. A replication incompetent retroviral lineage trace would directly examine clonal expansion. Furthermore, it would allow the authors to determine how many times a particular RPC divides. This relates to this idea of the mutant RPCs as being "mitotically old". How many divisions does it take for an RPC to reach this point? Why would reaching a mitotic limit and then producing a post-mitotic MG result in a degenerative MG? Might the death of the MGs that is observed at P14 and P30 simply be a pruning mechanism to rid the retina of excessive MGs that are not needed?

Figure 5.

Why didn't the authors do the Tsc1/Raptor double CKO and assess for rescue? This seems to be the most direct approach and implicate mTORC1 loss in the primary phenotype. The Hif1a CKO rescue may be indirect.

The Psmb-TKO rescue of the Tsc1 CKO phenotype is curious. Why would only a quarter rescue? Might the extensive crosses to generate this mouse have resulted in the segregation of a deleterious allele linked to the Trp1-Cre or different mice with 1 or 2 copies of Trp1-Cre?

Figure S6.

There is no indication of statistical significance.

Figure S11.

Glycolysis is the predominant mode of ATP synthesis in proliferative neural progenitor cells. Therefore, it stands to reason that lactate production was increased in the Tcs1 CKOs which the authors claim has accelerated cell cycle progression. However, it's not clear whether this is a primary or secondary phenotype.

Reviewer #3:

This report is a follow-up from an earlier study, in which the same lab reported Tsc1 cko leads to accelerated cell cycle progression. In this study, they repeat some of the same experiments, but they now show the partial rescue of the accelerated cell cycle progression with the concomitant cko of Hif1a (Figure 6), the Muller glial degeneration and retinal rosetting as the animals mature (Figure 1) and the dependence of the phenotype on the different cre-lines (Figure 4). These are all interesting phenotypes/results, but there are also some concerns, listed below.

1. The evidence for selective Muller glial cell death is not that strong. The demonstration of active Caspase in Muller glia in the supplement was not clear. Additional/better examples are needed. In addition, while rosettes can result from the loss of Muller glia, they can also result from increased or prolonged Muller glial proliferation (eg. p27 ko) or from inhibition of the BMP pathway during development, among other things. They use only Sox9 to label the Muller glia and the disruption in normal retinal histology might make accurate quantitation difficult. Additional markers and better characterization of the glial cells prior to their overt loss (earlier time points) would help understand the phenotype.

2. The evidence that the developmental over-proliferation of the Tsc1 cko progenitors and the Muller glial degeneration are linked is partly supported by the fact that deletion of Tsc1 in mature Muller glia did not cause this phenotype. However, it is not clear that the Tsc1 was effectively deleted in this experiment. More evidence to show this experiment actually produced the deletion should be provided.

3. One of the most interesting aspects of the report is that only the deletion of Tsc1 using the Trp1-cre line leads to the phenotypes. One interpretation of this is that the hyper-proliferation of the Tsc1 cko progenitors only occurs when there are normal cells nearby. The α-Pax6cre line shows a reduced phenotype, consistent with this idea, while when Tsc1 is knocked out in all progenitors across the retina, proliferation appears to be normal. They interpret these different phenotypes in terms of the "age" of the progenitors or the number of total cell divisions, but it seems more consistent with a model where normal and Tsc1 deficient progenitors compete for space and the Tsc1 deficient cells outcompete normal cells, but not other Tsc1 deficient progenitors. This alternative model could be tested by looking more closely at the percentages of EdU+ cells near the boundary of the conditional deletion. Probably best to do this early in the α-Pax6cre line where the boundary is likely to be sharp.

4. The neurosphere assay is not very quantitative to show differences between the Tsc1 cko (Figure 4), particularly when they do not see differences until the spheres are grown for 4 weeks, long past when proliferation would normally have finished in the retina. The in vivo cell cycle analyses in Figure 7 do a better job anyway.

5. They rescue the rosetting phenotype by crossing the Tsc1 cko mice with a Hif1a deletion. The "hyper-expansion" phenotype is less well rescued (Figures 6B, 6D). Thus it seems like these two phenotypes might be un-related. The authors should discuss this possibility.

6. The authors argue that Muller glia undergo senescence because they are the last cell type generated and the progenitors have exceeded their mitotic limit.This is not exactly true. Muller glia are among the last cells generated, but rods and bipolar cells are also included in the last cell divisions. However, these other last generated cells do not seem to be undergoing cell death or senescence-related gene expression. This is specifically demonstrated in supplemental Figure S9. Rather it appears that there is a specific requirement for Tsc1 in Muller glia, perhaps near the end of neurogenesis, much like there is for p27kip. It would be interesting to determine whether the premature end of neurogenesis might lead to incomplete differentiation of the Muller glia.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for submitting your article "mTORC1-induced retinal progenitor cell overproliferation leads to accelerated mitotic aging and degeneration of descendent Müller glia" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Carlos Isales as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

1) Please discuss the potential reasons as of why Chx10-Cre null neurospheres divide better in vitro as indicated by reviewer 1.

2) Please discuss why Tcs1 CKOs MG (based on tdTomato Cre reporter expression) seem to be unable of undergoing reactive gliosis, as pointed out by reviewer 2.

3) Please change the statement on page 21 that "The MG are the last-born retinal cell type, arising around the first postnatal week in mice." As reviewer 3 indicate this is not correct. Also discuss why rod and bipolar that are generated around the same time as Müller glia do not undergo the same apoptosis.

4) Please also discuss possible peculiarities of the ciliary margin, as indicated by reviewer 3.

Reviewer #1:

The authors have added important new experiments. First, they showed that using a TRP1-Cre-ER system, in which they activate Cre at E9.5 with tamoxifen, they obtain similar results to that observed with the TRP1-Cre system. This result addresses the concerns of the other reviewers that the integration site of TRP1-Cre may have been contributing to the phenotype.

Second, they isolated E13 progenitors from mice in which Tsc1 was deleted using Chx10-Cre, which does not induce the over-proliferation/Muller glia degeneration phenotype in vivo, and assessed resultant neurosphere size and number. They compared Tsc1 null to Tsc1 heterozygous neurospheres. Interestingly, Tsc1 null progenitors generated ~2-fold more neurospheres, and the spheres were somewhat (~20%) larger than those from Tsc1 hets. They then degenerated, mimicking the effect seen in vivo when TRP1-Cre is used to knockout Tsc1. These data imply that the effects of the knockout are cell autonomous, and are not a consequence of cell competition.

I commend the authors for their extra work to address the concerns of the reviewers, and the paper should be published. However, I do feel there should be an addition to the Discussion. The neurosphere/Chx10-Cre result also suggests that there is something about the in vivo milieu of the mice lacking the proliferation phenotype (Chx10-Cre, Rx-Cre etc) that suppresses the proliferative effect of Tsc1 deletion. One possibility, raised in the author's rebuttal, is that the ratio of mutant to normal cells must be low for the proliferative defect to occur, implying either that normal cells stimulate the mutant cell division, or excess mutant cells inhibit over-proliferation. Against the idea that normal cells stimulate over-proliferation, they now show that Chx10-Cre null neurospheres divide better in vitro. That leaves the notion that high numbers of Tsc1 cells may inhibit their own proliferation. The Discussion is completely lacking any mention of this confusing and unsolved aspect of the data, nor does it even acknowledge the paradox. At the very least the authors could highlight the problem and state that the difference in the effects of various Cre models in vivo coupled with the neurosphere data in vitro is puzzling, and the mechanism as to why mutant progenitors must be present at a low frequency to over-proliferate is unresolved.

Reviewer #2:

The reviewers have satisfied most of my concerns.

However, the mutant GFAP immunofluorescence raises an interesting question. The authors clearly show the presence of residual SOX9+ MGs within the mutant retinae with lamination defects. However, these mutant retinae, which are clearly experiencing damage, do not upregulate GFAP. These data suggest that these presumably Tcs1 CKOs MG (based on tdTomato Cre reporter expression) are not capable of undergoing reactive gliosis. Is MG differentiation, or the ability to detect damage, or Gfap transcription compromised?

Reviewer #3:

The authors have done a very good job at addressing my previous concerns. However, I think the text needs a few changes for accuracy. The authors state on page 21 that "The Mg are the last-born retinal cell type, arising around the first postnatal week in mice." This is not correct. The MG are AMONG the last cells born in the retina, but rods and bipolar cells are included in two cell clones that contain Muller glia. Therefore, the authors should change this sentence accordingly. Moreover, since at least some of the rods and most of the bipolar cells have been generated in the KO mice by progenitors that have undergone as many rounds of division as those that generated the MG, one might presume their survival would be similarly affected. Yet they appear not to undergo the apoptosis that the MG are subject to in these mice. The authors should discuss this point in the Discussion.

The second point that could bear some elaboration is the possibility that there is something unique about the Trp-cre expressing cells at the retinal margin. The authors show in Figure 4 that there is a progressive decline in RPC overgrowth from peripheral to central retina when the cre is targeted to progressively more RPCs, but these different cre lines also have a relative decrease in the contribution of RPCs adjacent to the CM. It is possible that this phenotype is primarily due to expansion in the proliferation of a relatively small subset of RPCs, ones that may more closely resemble the CMZ cells of lower vertebrates. The authors should consider/discuss this possibility.

https://doi.org/10.7554/eLife.70079.sa1

Author response

Essential revisions:

There is general agreement that the authors report a fascinating phenotype. However, there are also several concerns that are listed in the individual reviewers' reports. The common and most pressing unresolved issues that the authors should experimentally address are the following:

We appreciate the editors and reviewers for evaluating our report positively. We also thank for the comments that could certainly improve the quality of our paper.

1) Is the Trp1-Cre line somehow contributing to the phenotype? This could be solved by the use of theTrp1-CreERT2 line or the injection so AAV-Cre. Extensive discussion of this issue is provided in the individual reviewers' assessment of the manuscript.

We have also been aware of the toxicity issue of TRP1-Cre. Thus, we compared the phenotypes of Tsc1(f/+);TRP1-Cre and Tsc1(f/f);TRP1-Cre littermate mice to exclude the possibility that TRP1-Cre causes the phenotypes. In addition, those mice were obtained from the breeding of Tsc1(f/f) and Tsc1(f/+);TRP1-Cre pairs. Therefore, all mice carrying TRP1-Cre transgene are heterozygous for TRP1-Cre, and this could have reduced toxicity of TRP1-Cre.

As the reviewer recommended, we analyzed the phenotypes of Tsc1(f/f);TRP1-CreERT2 mice, in which Tsc1 was deleted in the CM and RPE populations in tamoxifen-dependent manner. We found that deletion of Tsc1 in early embryonic stages by E9.5 results in MG degeneration and retinal rosettes in R26tdTom-positive clones in the mature retina whereas the deletion in later stages did not (Figure 4 —figure supplement 2). The results suggest that Tsc1 deletion in early CM progenitors is necessary to provide a chance for their descendent RPCs to divide exceeding the division limit.

2) Can the results be interpreted in terms of cell competition? The study does not provide sufficient support to the statement that telomere shortening is a "cell autonomous" effect.

We agree to the reviewers’ opinions on the conclusion that the fast-dividing Tsc1-deficient clones might win the competition with slowly-dividing WT clones for occupying a space in the retina. This might result in the over proliferation of Tsc1-deficient RPCs exceeding the division limit. Supporting this, telomere shortening occurred excessively in Tsc1-deficient retinal cells in comparison to the telomeres of their neighboring WT cells (Figure 4B). However, we could not find evidence that compensatory hyper expansion and telomere shortening occur in WT cells in Raptor(f/f);TRP1-Cre mouse retina, in which Raptor-deficient cells were depleted precociously (Figure 5H). These results suggest that the overproliferation, which leads to excessive telomere shortening, is an autonomous phenomenon of Tsc1-deficent cells but a general feature of RPCs, which are surrounded by slowly-dividing neighboring RPCs.

In the revision, we also co-cultured Tsc1-deficient RPCs and WT RPCs to form neurospheres in neighbor, respecting the reviewer’s suggestion. We found the neurospheres of Tsc1-deficient RPCs expand faster than WT neurospheres, which present in neighbor or in a separate space (Figure 4 —figure supplement 3A – 3C). We also found the neurospheres derived from Tsc1-deficient RPCs exhibit the characteristics of mitotic aging, including telomere shortening, senescence, and apoptosis, with or without WT neurospheres (Figure 4 —figure supplement 3D – 3F). Therefore, the results also suggest that the clonal expansion and mitotic aging are intrinsic properties of Tsc1-deficient RPCs, which just need a space to expand their clones in the retina.

Reviewer #1:

[…]

– It is interesting, but overlooked in the paper, that the telomere length in KO cells is not intrinsically shorter in KO cells (e.g. not in mRx-Cre KO or CHx10-Cre KO retina), but only in retinas with mixed wt and KO cells (significant in TRP1-Cre KO, and n.s. in α-Cre but mean is lower). That finding suggests that the telomere shortening may be dependent on the presence of neighboring WT cells. This could, for example, reflect "cell competition" between KO and WT cells, in which the elevated rates of division is due to greater fitness of KO cells, that is stimulated by their weaker neighbors.

We agree to the reviewer’s opinion on the conclusion of competition between KO and WT cells for a space in the retina. As we wrote in the text, the first paragraph of section starts by “MG degeneration is correlated… (page 9)”, the ratio of fast-dividing Tsc1-deficient RPCs to slowly-dividing WT RPCs is likely a critical factor that determines whether the Tsc1-deficient RPCs can over proliferate and expand their clones exceeding the division limit. Thus, the Tsc1-deficient RPCs should be present in minority in the retina to have a competitive position in clonal expansion in the limited space of the retina, as those are in Tsc1(f/f);TRP1-Cre mice. On the contrary, in the other Tsc1-cko mice, the neighboring RPCs of a Tsc1-deficient RPC are mostly Tsc1-deficient, too, thus the RPC cannot win the competition with its neighbor. This difference might allow Tsc1-deficient RPCs to over proliferate and exceed the division limit only in Tsc1(f/f);TRP1-Cre mouse retina.

Indeed, mTOR signaling has been linked to cell competition in mammals: PMID: 29720666; PMID: 33652024. It should be feasible to assess this notion by isolating early stage KO cells and assessing whether their division rate and/or telomere attrition is boosted specifically in the presence of WT cells relative to pure KO cultures.

Respecting the reviewer’s suggestion, we co-cultured Tsc1-deficient RPCs and WT RPCs to form neurospheres in neighbor. We found the neurospheres of Tsc1-deficient RPCs expand faster than the neighboring WT neurospheres (Figure 4 —figure supplement 3A – 3C). We also found the neurospheres derived from Tsc1-deficient RPCs exhibit the characteristics of mitotic aging, including telomere shortening, senescence, and apoptosis, with or without WT neurospheres (Figure 4 —figure supplement 3D – 3F). Therefore, the results also suggest that the clonal expansion and mitotic aging are intrinsic properties of Tsc1-deficient RPCs. This idea was also supported by the finding that WT RPCs did not over proliferate to show extra telomere shortening in Raptor-cko mouse retina, where Raptor-deficient RPCs are depleted precociously.

– After showing telomere shortening in the KO but not wt cells from the same TRP1-Cre retina, the papers states that this phenomenon must be "cell autonomous". If that is the case, why is there no telomere shortening in other strains in which the entire/most of the retia is KO? As noted above, it appears that the shortening may depend on WT neighbors, in which case the term "cell autonomous" would be inappropriate.

We think the space where an RPC can expand is likely a limiting factor for clonal expansion. For a Tsc1-deficient RPC in Tsc1(f/f);TRP1-Cre mouse retinas, majority neighboring RPCs are slowly-expanding WT RPCs, thus they could have a competitively dominant position to expand fast in a retinal space. However, this chance for a Tsc1-deficient RPC should be much lower in the other Tsc1-cko mouse retinas, since its neighbors are also mostly fast dividing Tsc1-deficient RPCs. Thus, the Tsc1-deficient RPC cannot expand faster than its neighbors in other Tsc1-cko mouse retinas.

– The same is true of other biochemical phenotypes that are unique to the TRP1-Cre retina, notably senescence. It's feasible, therefore, that the entire basis for the phenotypic specificity in the TRP1-Cre retina is dependent on mutant/normal cell interaction. The authors should attempt assays to validate or rule out that possibility. In addition to the cell mixing studies above, it should also be feasible to perform clonal studies in explants or in vivo using retroviral Cre delivery. It would, for example, be feasible to at least assess whether the resulting clones contain senescent cells (e.g. SA-betagal-positive staining).

Owing to technical limitations, we could not deliver the virus into the eyes of early mouse embryos in the uterus. Instead, following the reviewer #2’s suggestion, we employed TRP1-CreERT2 mice to delete Tsc1 sparsely in the retina. We could find the degeneration of MG and rosette formation in the CreER-affected areas in Tsc1(f/f);TRP1-CreER mouse retinas (Figure 4 —figure supplement 2).

– The authors show that expressing Tert does not rescue elevated cell death in the TRP1-Cre KO retina. They conclude on page 12: "This suggests that telomere shortening is unlikely to be the cause of MG degeneration in Tsc1fl/fl;TRP1-Cre mice". They would need to demonstrate that telomere length is rescued to make that claim, otherwise they should modify to state: "This may mean either that Tert expression is insufficient to rescue telomere length in this context, or that telomere shortening does not cause MG degeneration"

We measured the telomere length of P30 dTert-TG and Tsc1-cko;dTert-TG mouse retinas, and found they have longer telomeres than WT and Tsc1-cko mice, respectively (Figure 5 —figure supplement 2C). The results suggest that telomere shortening is unlikely a direct cause of the phenotypes. We also modified the expression as “This suggests that telomere shortening is not a direct cause of MG degeneration in Tsc1fl/fl;TRP1-Cre mice”.

Overall, the paper is really interesting, but the authors have overlooked an explanation for their results that should be addressed, given the prominent literature on cell competition, and its link to mTOR.

Reviewer #2:

[…]

The main, unresolved conundrum is the lack of RPC phenotype when using the Chx10-Cre, mRx-Cre, and aPax6-Cre lines. I understand the authors' argument that since Trp1-Cre is active in fewer RPCs, the CKOs cells may be allowed to expand into the territory of the more slowly dividing, centrally located WT cells. However, if this is indeed the case, why was it not observed in the Chx10-Cre or aPax6-Cre mice? My lab and others have used these line for years and we can state with absolute certainty that the Cres are mosaically expressed. Therefore, one would definitely end up with patches of CKOs cells adjacent to WT cells. This is particularly true for aPax6-Cre that is predominantly expressed in peripheral RPCs. Wouldn't the CKO patches overtake the WT patches centrally?

All Tsc1-cko mice, which have the mosaicism of Cre activity, showed the expansion of Tsc1-deficient clones in their retinas (Figure 4, A – C), however only the deletion by TRP1-Cre resulted in the MG degeneration and retinal rosettes. We think the chance for a Tsc1-deficient RPC to expand its clone is dependent of the expansion power of neighboring RPCs. For a Tsc1-deficient RPC in Tsc1(f/f);TRP1-Cre mouse retinas, majority neighboring RPCs are slowly expanding WT RPCs, thus it could have a strong competitive position to expand among those WT RPCs. The WT cells unaffected by aPax6-Cre and Chx10-Cre are, however, less than 40% and 20%, respectively, thus majority neighbors of a Tsc1-deficient RPC are also Tsc1-deficient RPCs. Thus, the Tsc1-deficient RPC could not win a competition with neighboring Tsc1-deficent RPCs to expand in a retinal space.

The most parsimonious explanation for the discrepancies between the Trp1-Cre and the other RPC Cre lines, is that something other than Tsc1 CKO is contributing to the phenotype. I have several concerns regarding the Trp1-Cre line. Thanos, et al. (2012) reports that the line exhibits Cre toxicity leading to "RPE dysfunction and concomitant disorganization of RPE layer morphology, large areas of RPE atrophy, retinal photoreceptor dysfunction, and microglial cell activation in the affected areas". While the authors seem to be addressing this issue by comparing Trp1-Cre; Tsc1+/flox mice to Trp1-Cre; Tsc1flox/flox mice, are they certain that the mice in question all carry one copy of the Trp1-Cre transgene? This line is a random transgene insertion. Therefore, if one does not know where the transgene inserted into the genome, it's impossible to genotype for 1 versus 2 copies and a test cross would have to be used to determine zygosity. Different dosages of Cre could impact toxicity effects. Also, if the transgene inserted into an essential gene that could also lead to a phenotype in a homozygous state.

We have also been aware of the toxicity issue of TRP1-Cre. Thus, we compared the phenotypes of Tsc1(f/+);TRP1-Cre and Tsc1(f/f);TRP1-Cre littermate mice to exclude the possibility that TRP1-Cre causes the phenotypes. In addition, those mice were obtained from the breeding of Tsc1(f/f) and Tsc1(f/+);TRP1-Cre pairs. Therefore, all mice carrying TRP1-Cre transgene are heterozygous for TRP1-Cre, and this could have reduced toxicity of TRP1-Cre.

The authors should also consider the genetic background of these mice as several rd alleles are known to be present within certain inbred strains (i.e. rd8 in C57BL/6N).

We believe the genetic background of the mice converged to C57BL/6J after repeated backcrossing (>10 generation). We have also confirmed the mice used for the study do not have rd mutations, such as rd1, rd10, and rd8 (data not shown), by genotyping.

This reviewer recognizes the tremendous effort put forth in this paper and sincerely appreciates it. The utilization of mouse genetics is truly impressive, and I commend you for that. Nevertheless, I still cannot shake the concern that there is something about the Trp1-Cre line that is misleading you. It's very odd to me that the other RPC Cre lines show absolutely no phenotype in vivo. I understand that this is not an easy question to answer, but it would benefit your paper greatly if you could show some direct evidence that the Trp1-Cre mediated phenotype is specific to Tcs1 loss in RPCs without any contribution from negative effects of the Trp1-Cre itself. It's formally possible that Trp1-Cre on its own has an effect that's synergizing with Tsc1 loss to give you the phenotype.

We have also been aware of the toxicity issue of TRP1-Cre. Thus, we compared the phenotypes of Tsc1(f/+);TRP1-Cre and Tsc1(f/f);TRP1-Cre littermate mice to exclude the possibility that TRP1-Cre causes the phenotypes. We have also confirmed the mice used for the study do not have rd mutations, such as rd1, rd10, and rd8 (data not shown), by genotyping.

You could employ the Trp1-CreERT2 line (Mori, et al., 2012). This an independent transgene and the random integration site is expected to be completely different from the Trp1-Cre transgene. Also, because Cre activity is tamoxifen-dependent, this line would be an effective means to confirm the Trp1-Cre Tsc1 CKO phenotype as being specific to Trp1-Cre mediated recombination of the floxed allele. In other words, the comparison would be Trp1-CreERT2; Tsc1flox/flox littermates with or without tamoxifen. However, there could still be issues with Cre toxicity upon induction. Another method could be AAV-Cre which could be used to sparsely infect RPCs with Cre followed by assessment of CKO cell expansion into WT areas. This is probably the preferred method as it does not depend on the Trp1 promoter.

As the reviewer recommended, we analyzed the phenotypes of Tsc1(f/f);TRP1-CreERT2 mice, in which Tsc1 was deleted in the CM and RPE populations in tamoxifen-dependent manner. We found that deletion of Tsc1 in early embryonic stages by E9.5 results in MG degeneration and retinal rosettes in R26tdTom-positive clones in the mature retina whereas the deletion in later stages did not (Figure 4 —figure supplement 2). The results suggest that Tsc1 deletion in early CM progenitors is necessary to provide a chance for their descendent RPCs to divide exceeding the division limit. However, owing to technical limitations, we could not deliver AAV-Cre virus into the eyes of early mouse embryos across the uterus.

Figure S1.

Here the authors show pS6 immunofluorescence to indicate expression in the retina but not the ciliary margin or RPE. Why is the retina staining so sparse? Wouldn't one expect more widespread expression with the proliferative RPC population? What is the identity of the pS6+ cells?

Previous reports have also shown the sparse pS6 signals in developing mouse retina (Choi et al., 2018; Hagglund et al., 2017). The pS6 signals were detected in RPCs (Vsx2-positive) and PMNs (Tuj1-positive) (Figure 1 —figure supplement 1). The former pS6 signals might be related with the acceleration of cell cycle, while the latter might be related with the growth and maturation of PMNs. The mTORC1 activity can be changed dynamically during cell cycle, thus pS6 might be detectable at high in sub-RPC population at certain cell cycle.

Figure 1.

Is the ciliary body actually "absent" as the authors state or is it malformed?

We examined the distribution of CB markers, such as Aqp1 and Otx1, in the mouse eye sections. The Aqp1- and Otx1-positive cells were detectable in reduced numbers in the peripheral parts of Tsc1(f/f);TRP1-Cre mouse eyes, suggesting the hypoplasia of CB (please see Author response image 1). We, thus, modified the expression that “the ciliary body is malformed”, in the revised text.

Author response image 1
CB malformation in Tsc1f/f;TRP1-Cre mouse eyes.

(A) Distribution of CB cells in the littermate mouse eyes was investigated by immunostaining of specific markers, such as Otx1 and Aqp1. (B) Numbers of Otx1-stained nuclei in the peripheral part of the eye sections were counted and relative numbers against Tsc1f/+;TRP1-Cre samples are shown in the graph. **, p<0.01; ***, p<0.001.

At P14 and P30, if MGs are missing due to apoptosis, why are other retinal cell types that require MGs for homeostatic support also not degenerative?

Other retinal cell types also degenerated later in the Tsc1(f/f);TRP1-Cre mouse retinas (please see Author response image 2). However, those retinal cells were not affected by P14, when MG loss was already evident (Figure 1D; Figure1 —figure supplement 2A). The results suggest that the degeneration of retinal neurons might start when the structural and functional changes caused by MG loss are accumulated in the retina.

Author response image 2
Degeneration of retinal neurons in Tsc1f/f;TRP1-Cre adult mice.

(A) Distribution of Cre-affected cells in 9 months-old littermate mouse retinas are visualized by R26tdTom reporter. Activation of mTORC1 in the retinas were determined by pS6 immunostaining. (B) Distribution of retinal cell type-specific markers (explained in Figure 1 and Figure 1 —figure supplement 2) in the retinas were also investigated by immunostaining.

Might the MG loss actually just be a downregulation of Sox9 and GS expression during gliosis? GFAP, Cyclin D1, and p27 immunofluorescence could address this. However, it's also curious that in Figure S2 the authors don't show an increase in GFAP expression in the MG of the CKO mice. Such disruptions in retinal architecture would certainly be expected to result in a gliotic response of whatever MGs are still present near the affected area.

We provide the immunostaining results of other MG markers, such as p27 and Sox2, in the revised Figure 1 —figure supplement 2A. The quantifications of the results are also provided in Figure 1D.

We also co-stained Gfap and Sox9 to determine the gliotic responses of MG in mouse retina at P30, when Gfap-positive signals are observed in Tsc1(f/f);TRP1-Cre mouse inner retina (Figure 1 —figure supplement 2A). However, those inner retinal Gfap signals are mostly Sox9-negative (please see author response image 3), suggesting that those are not MG but might be the extension of astrocyte cell processes.

Author response image 3
Activation and gliosis of MG in the mouse retinas were determined by co-staining of Sox9, a MG marker, and Gfap, which is expressed in astrocytes and activated MG.

Figure 3.

The Sox9+ MG layer in the Chx10-Cre; Tsc1flox/flox mice looks very disorganized. Is this a representative image? If it is, it might reflect a subtle phenotype. The MG layer sometimes appears less organized when undergoing a gliotic response.

Nuclear positions of the INL cells are disorganized in Tsc1-cko mouse retinas because of the cytomegaly and excessive branching of Tsc1-deficient retinal neurons and MG (Figure 1C; Figure1 —figure supplement 2A). Similar results have also been reported previously (Choi et al. (2018); Hagglund et al. (2017)).

Figure 4.

Please indicate statistical significance in panels E, G, and H.

We added the p-values in the graphs.

I take issue with the term "clonal expansion" as that's not really what this experiment examines. A replication incompetent retroviral lineage trace would directly examine clonal expansion. Furthermore, it would allow the authors to determine how many times a particular RPC divides.

It is necessary to inject the virus at low copy into the early mouse embryonic mouse eyes across the uterus. Owing to a technical limitation, we could not deliver a replication incompetent retrovirus into early mouse embryonic eyes. We, instead, used TRP1-CreERT2, to address this point (please see our response to your major comment above).

This relates to this idea of the mutant RPCs as being "mitotically old". How many divisions does it take for an RPC to reach this point?

Unfortunately, we do not know the exact number of mouse RPC division limit, which should be counted from one cell stage of the embryo. Instead, we assessed relative division numbers by calculating the cells comprising a neurosphere that reaches the maximum number (Figure 4F – 4G; Figure 4 —figure supplement 3).

Why would reaching a mitotic limit and then producing a post-mitotic MG result in a degenerative MG? Might the death of the MGs that is observed at P14 and P30 simply be a pruning mechanism to rid the retina of excessive MGs that are not needed?

Excessive retinal cells produced during development are known to degenerate in the postnatal stages as the reviewer indicates. This developmental pruning leaves the cells in constant numbers at the end. The numbers of MG were much less than the normal in the mature Tsc1(f/f);TRP1-Cre mouse retinas in comparison to their littermates’ (i.e., P14 and P30), whereas they were more in the developing retina (i.e., P7) (Figure 1D; Figure 1 —figure supplement 2). The numbers of other retinal cell types were, however, not changed significantly, although they were also produced at higher numbers ahead of their regular schedules (Choi et al., 2018; Figure 1D; Figure 1 —figure supplement 2). Therefore, we think the loss of MG in the Tsc1(f/f);TRP1-Cre mouse retina is unlikely to result from excessive pruning.

Figure 5.

Why didn't the authors do the Tsc1/Raptor double CKO and assess for rescue? This seems to be the most direct approach and implicate mTORC1 loss in the primary phenotype. The Hif1a CKO rescue may be indirect.

The phenotypes of Tsc1/Raptor double cko mouse retinas were not greatly different from those of Raptor-cko mouse retinas. We provide the results in Author response image 4.

Author response image 4
Eye and retinal structures of P30 littermate mice with indicated genotypes were investigated by H&E staining.

Distribution of the Cre-affected cells and mTORC1 activation of the cells were examined by the immunostaining of R26EYFP and pS6, respectively.

The Psmb-TKO rescue of the Tsc1 CKO phenotype is curious. Why would only a quarter rescue?

We think the incomplete rescue might be related with the adaptation of RPCs, which might utilize alternative proteasome machineries in the absence of the immunoproteasome and over proliferate. The alternative proteasomes were, however, not able to compensate the immunoproteasome in a quarter, thus the phenotypes cannot appear in their retinas.

Might the extensive crosses to generate this mouse have resulted in the segregation of a deleterious allele linked to the Trp1-Cre or different mice with 1 or 2 copies of Trp1-Cre?

The mice were obtained from the breeding of Tsc1(f/f) and Tsc1(f/+);TRP1-Cre pairs. Therefore, all mice carrying TRP1-Cre allele are TRP1-Cre(+/-).

Figure S6.

There is no indication of statistical significance.

We added the p-values in the graph.

Figure S11.

Glycolysis is the predominant mode of ATP synthesis in proliferative neural progenitor cells. Therefore, it stands to reason that lactate production was increased in the Tcs1 CKOs which the authors claim has accelerated cell cycle progression. However, it's not clear whether this is a primary or secondary phenotype.

We think the increase of lactate production is a secondary phenotype caused by the shortage of intracellular ATP, which was resulted from enhanced energy consumption of Tsc1-deficient RPCs.

Reviewer #3:

This report is a follow-up from an earlier study, in which the same lab reported Tsc1 cko leads to accelerated cell cycle progression. In this study, they repeat some of the same experiments, but they now show the partial rescue of the accelerated cell cycle progression with the concomitant cko of Hif1a (Figure 6), the Muller glial degeneration and retinal rosetting as the animals mature (Figure 1) and the dependence of the phenotype on the different cre-lines (Figure 4). These are all interesting phenotypes/results, but there are also some concerns, listed below.

1. The evidence for selective Muller glial cell death is not that strong. The demonstration of active Caspase in Muller glia in the supplement was not clear. Additional/better examples are needed.

We determined the apoptotic cell death of MG by TUNEL assay and immunostaining of active caspase-3 (Casp-3), which are generally used for the detection of apoptotic cells. In the original Figure S3, the apoptotic MG are, however, not identified clearly by Casp-3 immunostaining. Thus, we modified the protocol for the staining to visualize Casp-3 more clearly (please see revised Figure 1 —figure supplement 3A). In the modified staining condition, Sox9 was, however, not detectable, thus Sox2 was used to identify MG in the sections.

In addition, while rosettes can result from the loss of Muller glia, they can also result from increased or prolonged Muller glial proliferation (eg. p27 ko) or from inhibition of the BMP pathway during development, among other things.

We could not find EdU incorporation in GS-positive MG cells in P14 mouse retinas (please see Author response image 5), suggesting that there was no prolonged MG proliferation in the mouse retinas.

Author response image 5
Proliferation of MG in P14 littermate mouse retinas was determined by the presence of GS-positive MG, which incorporated EdU for 24h.

We could not find the changes of phosphorylated Smad1 and 5 (pSmad1/5), which reflect the activation of BMP signaling, in the mouse retinas (please see Author response image 6).

Author response image 6
Activity of BMP signaling pathway in mouse retinas were determined indirectly by detecting phosphorylated Smad 1 and 5 (pSmad1/5), which are induced by BMP-activated receptors.

They use only Sox9 to label the Muller glia and the disruption in normal retinal histology might make accurate quantitation difficult. Additional markers and better characterization of the glial cells prior to their overt loss (earlier time points) would help understand the phenotype.

We have also detected MG by additional markers, including GS, p27, and Sox2. We added the results in Figure 1D and Figure 1 —figure supplement 2A.

2. The evidence that the developmental over-proliferation of the Tsc1 cko progenitors and the Muller glial degeneration are linked is partly supported by the fact that deletion of Tsc1 in mature Muller glia did not cause this phenotype. However, it is not clear that the Tsc1 was effectively deleted in this experiment. More evidence to show this experiment actually produced the deletion should be provided.

The best way to determine the deletion of Tsc1 gene in Glast-CreER-affected MG might be examining Tsc1 expression in R26tdTom-positive cells by co-immunostaining. However, the anti-Tsc1 antibody did not work for immunostaining. Thus, instead, we detected Cre-dependent deletion in Tsc1 gene locus by PCR. Our results show that the floxed DNA, which includes exon 17 and 18 of mouse Tsc1 gene, was deleted only in Tsc1(f/f);Glast-CreER mouse retina injected with tamoxifen. The results suggest that Tsc1 was deleted in the Glast-CreER-affected mouse MG. The efficacy of Glast-CreER has also been confirmed in many previous reports, including the original paper (de Melo et al., 2012).

Author response image 7
PCR detection of Glast-CreER-mediated deletion in mouse Tsc1 gene locus.

(A) Schematic diagram of Tsc1 gene locus of wt and Tsc1flox mice. (B) Agarose gel images of DNA bands amplified by the PCR with indicated primers. 1. loxP containing DNA fragments of Tsc1 gene, which were not affected by Cre recombinase (495bp). 2. wt DNA fragments of Tsc1 gene (295bp). 3. 2.01kbp and 2.24kbp DNA fragments are expected to be amplified from wt and Tsc1flox alleles, respectively. However, those are too big to be amplified in our PCR condition (extension time = 60 seconds). 4. Tsc1 exon 17&18 were deleted in MG subpopulation by Glast-CreERt2 only upon Tam injection (368bp). 5. Tsc1 exon 17&18 were deleted in majority retinal cells by Chx10-Cre (368bp).

3. One of the most interesting aspects of the report is that only the deletion of Tsc1 using the Trp1-cre line leads to the phenotypes. One interpretation of this is that the hyper-proliferation of the Tsc1 cko progenitors only occurs when there are normal cells nearby. The α-Pax6cre line shows a reduced phenotype, consistent with this idea, while when Tsc1 is knocked out in all progenitors across the retina, proliferation appears to be normal. They interpret these different phenotypes in terms of the "age" of the progenitors or the number of total cell divisions, but it seems more consistent with a model where normal and Tsc1 deficient progenitors compete for space and the Tsc1 deficient cells outcompete normal cells, but not other Tsc1 deficient progenitors. This alternative model could be tested by looking more closely at the percentages of EdU+ cells near the boundary of the conditional deletion. Probably best to do this early in the α-Pax6cre line where the boundary is likely to be sharp.

We agree that the space allowance for a Tsc1-deficient RPC is also a critical factor that determines whether it can divide exceeding the division limit. To be a dominant clone in a space, a fast-dividing Tsc1-deficient RPC should have majority WT RPCs that divide slowly in its neighbor, as the case of Tsc1(f/f);TRP1-Cre mice. However, the neighbors of a Tsc1-deficient RPC are also mostly Tsc1-defcient RPCs in the other Tsc1-cko mice, thus the RPC cannot hyperexpand exceeding its neighbors.

We also compared EdU(+) cells in tdTom(+) Tsc1-deficient clones and adjacent tdTom(-) WT clones in P0 Tsc1(f/f);TRP1-Cre mouse retinas. We could find the increase of EdU-positivity in tdTom(+) clones in the mouse retina, suggesting the hyperproliferation of Tsc1-deficient RPCs over neighboring WT RPCs (please see the results Author response image 8). In our previous report of Tsc1-cko by Chx10-Cre (Choi et al., 2018), we had also compared the proliferation rate of R26(+) potential Tsc1-deficient cells and R26(-) WT cells in the same retina. The results showed enhanced cell proliferation in R26(+) areas in comparison to their adjacent R26(-) areas.

Author response image 8
Relative overproliferation of Tsc1-deficient RPCs in Tsc1f/f;TRP1-Cre mouse retina.

(A) EdU-labeled proliferating cells and those moved to G2/M phase of cell cycle to express pH3 in P0 mouse retinas were examined by co-immunostaining after the injection of EdU to the mice at 3 h prior to tissue preparation. (B) Numbers of EdU-labeled proliferating cells in R26tdTom(+) Cre-affected cell areas and R26tdTom(-) wild-type cell areas were counted and shown in the graph. (C) Numbers of EdU;pH3-positive cells in R26tdTom(+) Cre-affected areas and R26tdTom(-) wild-type cell areas were counted and shown in the graph.

4. The neurosphere assay is not very quantitative to show differences between the Tsc1 cko (Figure 4), particularly when they do not see differences until the spheres are grown for 4 weeks, long past when proliferation would normally have finished in the retina. The in vivo cell cycle analyses in Figure 7 do a better job anyway.

5. They rescue the rosetting phenotype by crossing the Tsc1 cko mice with a Hif1a deletion. The "hyper-expansion" phenotype is less well rescued (Figures 6B, 6D). Thus it seems like these two phenotypes might be un-related. The authors should discuss this possibility.

mTORC1 might regulate cell growth and proliferation via multiple downstream targets. Hif1alpha is one of the targets, therefore Hif1alpha deletion could not normalize completely the Tsc1-cko phenotypes. Besides Hif1alpha, the hyperproliferation of Tsc1-deficient RPCs might be mediated by multiple mTORC1 targets including the immunoproteasome, of which loss was also insufficient to rescue Tsc1-cko phenotypes (Figure 6 —figure supplement 1). However, Hif1a deletion is enough to reduce the Tsc1-deficient clone size below the threshold level, which is necessary for the degeneration of MG.

6. The authors argue that Muller glia undergo senescence because they are the last cell type generated and the progenitors have exceeded their mitotic limit. This is not exactly true. Muller glia are among the last cells generated, but rods and bipolar cells are also included in the last cell divisions. However, these other last generated cells do not seem to be undergoing cell death or senescence-related gene expression. This is specifically demonstrated in supplemental Figure S9. Rather it appears that there is a specific requirement for Tsc1 in Muller glia, perhaps near the end of neurogenesis, much like there is for p27kip. It would be interesting to determine whether the premature end of neurogenesis might lead to incomplete differentiation of the Muller glia.

Thank you for the comment that provides us an alternative interpretation of the phenotype. In this study, we have not determined whether MG cells differentiate completely or not. Only mature MG marker we examined in this study is GS, which was expressed properly in Tsc1(f/f);TRP1-Cre as well as Tsc1(f/+);TRP1-Cre mice (Figure 1C). The GS-positive cells were even produced at higher number in P7 Tsc1(f/f);TRP1-Cre mouse retinas in comparison to those in Tsc1(f/+);TRP1-Cre littermate retinas (Figure 1C; Figure 1 —figure supplement 2B). However, the correct answer for the question could be provided by comprehensive analyses of gene expression in MG of the mouse retinas. The analyses may need scRNA-seq or RNA-seq of purified MG in the mouse retinas. Thus, it should be done in a separate study in future.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Essential revisions:

1) Please discuss the potential reasons as of why Chx10-Cre null neurospheres divide better in vitro as indicated by reviewer 1.

We think that our explanation for the conclusion might not be enough in the previous versions. Our conclusion that Tsc1-deficient RPCs can have the highest over proliferation potential when they are in minority (i.e., the deletion by Tyrp1-Cre) does not emphasize the importance of non-autonomous factors provided by neighboring wild-type cells. A space that is not occupied by wild-type clones can be the factor, if there is a factor contributed by the neighboring wild-type cells.

As the reviewer indicates, our neurosphere results show the exhaustive over proliferation can also happen to Chx10-Cre Tsc1-deficient RPCs (Figure 5F – 5H; Figure 5 —figure supplement 3F), which did not expand exceeding the limit in vivo (Figure 5A and 5C). One important difference of the RPCs in the neurosphere culture from those in vivo is a space to expand their clones. The space allowed for a Chx10-Cre Tsc1-deficient RPC in the neurosphere culture is enough to expand until they reach division limit, whereas the space is limited in mouse retinas. Furthermore, the Tsc1-deficient RPC has to compete with another Tsc1-deficient RPCs (>80% of neighbors) for a retinal space in Tsc1(fl/fl);Chx10-Cre mouse retinas. This chance to have another Tsc1-deficient RPCs in the neighbor is much lower in Tsc1(fl/fl);Tyrp1-Cre mouse retina (<50% of neighbors). Therefore, a Tsc1-deficient RPC could expand more freely in Tsc1(fl/fl);Tyrp1-Cre mouse retina than the other three Tsc1-cko mouse retinas (i.e., Rax-Cre, Chx10-Cre, and α-Cre).

We modified the conclusion by reflecting these points (page 9 and 21). To help the readers’ understanding the data, we also provide a hypothetical model diagram in Figure 4 —figure supplement 4.

2) Please discuss why Tcs1 CKOs MG (based on tdTomato Cre reporter expression) seem to be unable of undergoing reactive gliosis, as pointed out by reviewer 2.

Currently, we cannot provide the correct answer why the MG derived from Tsc1-deficient RPCs do not exhibit the characteristics of reactive gliosis. The Tsc1(fl/fl);Tyrp1-Cre mouse retina exhibited the increases of Iba1-positive microglia cell number and the hyper-extension of astrocyte processes (please see the results in the following response to the reviewer’s comment), indicating that the damages are present in the retina. These results also suggest that mTORC1 hyperactivation in Tsc1-deficient MG might suppress the gliotic responses. The mechanism for this possibility, however, should be investigated in a future study.

3) Please change the statement on page 21 that "The MG are the last-born retinal cell type, arising around the first postnatal week in mice." As reviewer 3 indicate this is not correct. Also discuss why rod and bipolar that are generated around the same time as Müller glia do not undergo the same apoptosis.

We appreciate for the correction. We corrected it in the revised manuscript (highlighted in page 21).

4) Please also discuss possible peculiarities of the ciliary margin, as indicated by reviewer 3.

In the revised Discussion (page 22), we added our interpretation why the effects of Tsc1 deletion could be greater in the CM RPCs than the central RPCs.

Reviewer #1:

The authors have added important new experiments. First, they showed that using a TRP1-Cre-ER system, in which they activate Cre at E9.5 with tamoxifen, they obtain similar results to that observed with the TRP1-Cre system. This result addresses the concerns of the other reviewers that the integration site of TRP1-Cre may have been contributing to the phenotype.

Second, they isolated E13 progenitors from mice in which Tsc1 was deleted using Chx10-Cre, which does not induce the over-proliferation/Muller glia degeneration phenotype in vivo, and assessed resultant neurosphere size and number. They compared Tsc1 null to Tsc1 heterozygous neurospheres. Interestingly, Tsc1 null progenitors generated ~2-fold more neurospheres, and the spheres were somewhat (~20%) larger than those from Tsc1 hets. They then degenerated, mimicking the effect seen in vivo when TRP1-Cre is used to knockout Tsc1. These data imply that the effects of the knockout are cell autonomous, and are not a consequence of cell competition.

I commend the authors for their extra work to address the concerns of the reviewers, and the paper should be published. However, I do feel there should be an addition to the Discussion. The neurosphere/Chx10-Cre result also suggests that there is something about the in vivo milieu of the mice lacking the proliferation phenotype (Chx10-Cre, Rx-Cre etc) that suppresses the proliferative effect of Tsc1 deletion. One possibility, raised in the author's rebuttal, is that the ratio of mutant to normal cells must be low for the proliferative defect to occur, implying either that normal cells stimulate the mutant cell division, or excess mutant cells inhibit over-proliferation.

We think that our explanation for the conclusion might not be enough in the previous versions. Our conclusion does not emphasize the importance of non-autonomous factors provided by majority neighboring wild-type cells for the over proliferation of Tsc1-deficient RPCs. A space that is not occupied by wild-type clones can be the factor, if there is a factor contributed by the neighboring wild-type cells. Please see the details in our response below.

Against the idea that normal cells stimulate over-proliferation, they now show that Chx10-Cre null neurospheres divide better in vitro. That leaves the notion that high numbers of Tsc1 cells may inhibit their own proliferation. The Discussion is completely lacking any mention of this confusing and unsolved aspect of the data, nor does it even acknowledge the paradox. At the very least the authors could highlight the problem and state that the difference in the effects of various Cre models in vivo coupled with the neurosphere data in vitro is puzzling, and the mechanism as to why mutant progenitors must be present at a low frequency to over-proliferate is unresolved.

As we show in vitro neurosphere culture (Figure 4, F – H and figure supplement 3) and EdU-labeling in vivo (Figure 7, A – C), Tsc1-deficient RPCs are intrinsically capable of over proliferation regardless of the Cre drivers. The exhaustive proliferation, however, can occur only if the spatial limitation is not present.

As we show in the hypothetical model diagram in Figure 4 —figure supplement 4, by having more wild-type RPCs in its neighbor, a Tsc1-deficient RPC in Tsc1(fl/fl);Tyrp1-Cre mouse retina can have a higher chance than that in Tsc1(fl/fl);Chx10-Cre mouse retinas to divide extra rounds and fill a free space, which neighboring wild-type RPCs did not fill yet. This invasive clonal expansion will continue until the Tsc1-deficient RPCs reach the division limit and cannot divide more. Based on these, we concluded that Tsc1-deficient RPC population should be lower than a threshold level that a Tsc1-deficient RPC clone can escape the competition with another Tsc1-deficient RPC and over proliferate exceeding the division limit.

This type of competitive clonal expansion could also occur in other conditions, where two RPC populations expand at different speeds. The Rptor(fl/fl);Tyrp1-Cre mouse retina, which is composed of relatively fast dividing wild-type RPCs and slowly dividing (or cell cycle arrested) Rptor-deficient RPCs, could be also one of the examples. However, wild-type RPCs in the Rptor(fl/fl);Tyrp1-Cre mouse retina could not over proliferate even though their neighboring Rptor-deficient RPCs stopped expanding (Figure 5F – 5J). The results suggest that the over proliferation is not a relative property obtained by the division differences but a specific property of Tsc1-deficient RPC.

Combining these, we modified our interpretation for the clonal hyper-expansion in Discussion (second paragraph in page 21). We also provide a model diagram that shows the hypothetical RPC expansion in each Tsc1-cko and Rptor-cko mouse retinas (Figure 4 —figure supplement 4).

Reviewer #2:

The reviewers have satisfied most of my concerns.

However, the mutant GFAP immunofluorescence raises an interesting question. The authors clearly show the presence of residual SOX9+ MGs within the mutant retinae with lamination defects. However, these mutant retinae, which are clearly experiencing damage, do not upregulate GFAP. These data suggest that these presumably Tcs1 CKOs MG (based on tdTomato Cre reporter expression) are not capable of undergoing reactive gliosis. Is MG differentiation, or the ability to detect damage, or Gfap transcription compromised?

(1) Regarding to MG differentiation: Given the expression of GS (Figure 1C; Figure 1 —figure supplement 2B), a marker for mature MG, we believe MG are likely differentiated properly in the Tsc1(fl/fl);Tyrp1-Cre mouse retina.

(2) Regarding to the retinal damage: Tsc1(fl/fl);Tyrp1-Cre mouse retina exhibited the increase of Iba1-positive microglia cell number and the hyper-extension of astrocyte processes into the retina (please see the results in Author response image 9). These results suggest the presence of the damages in the retina. However, we do not know whether the damage sensing mechanism of Tsc1-deficient MG is compromised or not.

Author response image 9
Distribution of glia in the mouse retina.

(3) Regarding to Gfap expression: It would be possible that hyperactive mTORC1 suppresses the expression of Gfap in the MG. However, we do not know whether Gfap transcription is compromise until we assess the transcription efficiency at the Gfap locus. We would like to leave this for a future work.

Reviewer #3:

The authors have done a very good job at addressing my previous concerns. However, I think the text needs a few changes for accuracy. The authors state on page 21 that "The Mg are the last-born retinal cell type, arising around the first postnatal week in mice." This is not correct. The MG are AMONG the last cells born in the retina, but rods and bipolar cells are included in two cell clones that contain Muller glia. Therefore, the authors should change this sentence accordingly.

Thank you for the correction. We corrected and highlighted it in the manuscript (page 21).

Moreover, since at least some of the rods and most of the bipolar cells have been generated in the KO mice by progenitors that have undergone as many rounds of division as those that generated the MG, one might presume their survival would be similarly affected. Yet they appear not to undergo the apoptosis that the MG are subject to in these mice. The authors should discuss this point in the Discussion.

MG might be in less terminal stage than the other last-born types (i.e., rPR and BC) in terms of differentiation, since they can resume cell cycle to regenerate neurons in the injured cold-blooded vertebrate retinas (reviewed by Lahne et al., 2020). Mouse MG could also divide after the injury, if histone deacetylase (Hdac) inhibitor is provided after viral expression of a proneural transcription factor achaete-scute homolog 1 (Ascl1) (Jorstad et al., 2017). Furthermore, given the fact that developing cells are more sensitive than terminally differentiated cells to cell death (Fuchs and Steller, 2011; Vaux and Korsmeyer, 1999), MG therefore are likely more sensitive to the cell death than rPR and BC. We added this interpretation with the references in Discussion (page 22 and 23).

The second point that could bear some elaboration is the possibility that there is something unique about the Trp-cre expressing cells at the retinal margin. The authors show in Figure 4 that there is a progressive decline in RPC overgrowth from peripheral to central retina when the cre is targeted to progressively more RPCs, but these different cre lines also have a relative decrease in the contribution of RPCs adjacent to the CM. It is possible that this phenotype is primarily due to expansion in the proliferation of a relatively small subset of RPCs, ones that may more closely resemble the CMZ cells of lower vertebrates. The authors should consider/discuss this possibility.

We agree that the deletion of Tsc1 in a small RPC subset, like those in the CM, leads to the phenotypes through the competitive over proliferation of the Tsc1-deficient RPCs against majority wild-type RPCs. The CM RPCs were found to proliferate less robustly than those in the central retina in mouse embryo (Bélanger et al., 2017; Marcucci et al., 2016), so did the RPCs in lower vertebrate CMZ (Harris and Perron, 1998). Therefore, the effects of Tsc1 deletion that accelerates RPC cell cycle might be greater in the CM RPCs than the majority central RPCs. We added this interpretation with the references in Discussion (page 21).

https://doi.org/10.7554/eLife.70079.sa2

Article and author information

Author details

  1. Soyeon Lim

    Department of Biological Sciences, Korea Advanced Institute of Science and Technology, Daejeon, Republic of Korea
    Contribution
    Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Writing – original draft
    Competing interests
    No competing interests declared
  2. You-Joung Kim

    Department of Biological Sciences, Korea Advanced Institute of Science and Technology, Daejeon, Republic of Korea
    Contribution
    Formal analysis, Investigation, Methodology
    Competing interests
    No competing interests declared
  3. Sooyeon Park

    Department of Biological Sciences, Korea Advanced Institute of Science and Technology, Daejeon, Republic of Korea
    Contribution
    Methodology, Resources
    Competing interests
    No competing interests declared
  4. Ji-heon Choi

    Department of Biological Sciences, Korea Advanced Institute of Science and Technology, Daejeon, Republic of Korea
    Contribution
    Methodology, Resources
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-9204-1755
  5. Young Hoon Sung

    1. Department of Biochemistry, College of Life Science and Biotechnology, Yonsei University, Yonsei, Republic of Korea
    2. Department of Convergence Medicine, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Republic of Korea
    Contribution
    Methodology, Resources
    Competing interests
    No competing interests declared
  6. Katsuhiko Nishimori

    Department of Obesity and Internal Inflammation; Bioregulation and Pharmacological Medicine, Fukushima Medical University, Fukushima, Japan
    Contribution
    Resources
    Competing interests
    No competing interests declared
  7. Zbynek Kozmik

    Institute of Molecular Genetics of the Czech Academy of Sciences, Prague, Czech Republic
    Contribution
    Resources
    Competing interests
    No competing interests declared
  8. Han-Woong Lee

    Department of Biochemistry, College of Life Science and Biotechnology, Yonsei University, Yonsei, Republic of Korea
    Contribution
    Methodology, Resources
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-9515-3605
  9. Jin Woo Kim

    Department of Biological Sciences, Korea Advanced Institute of Science and Technology, Daejeon, Republic of Korea
    Contribution
    Conceptualization, Data curation, Funding acquisition, Project administration, Supervision, Writing – original draft, Writing – review and editing
    For correspondence
    jinwookim@kaist.ac.kr
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-0767-1918

Funding

National Research Foundation of Korea (2017R1A2B3002862)

  • Jin Woo Kim

National Research Foundation of Korea (2018R1A5A1024261)

  • Jin Woo Kim

Samsung Science and Technology Foundation (SSTF-BA1802-10)

  • Jin Woo Kim

Czech Science Foundation (21-27364S)

  • Zbynek Kozmik

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

Acknowledgements

This work was supported by the National Research Foundation of Korea (NRF) grants funded by Korean Ministry of Science and ICT (MSIT) (2017R1A2B3002862 and 2018R1A5A1024261; JWK); the grant funded by Samsung Foundation of Science and Technology (SSTF-BA1802-10; JWK); and the grant of Czech Science Foundation (21–27364S).

Ethics

Experiments using the mice were carried out according to the guidance of Institutional Animal Care and Use Committee (IACUC) of KAIST (KA-2014-20).

Senior Editor

  1. Carlos Isales, Medical College of Georgia at Augusta University, United States

Reviewing Editor

  1. Paola Bovolenta, CSIC-UAM, Spain

Publication history

  1. Received: May 5, 2021
  2. Accepted: October 17, 2021
  3. Accepted Manuscript published: October 22, 2021 (version 1)
  4. Version of Record published: November 9, 2021 (version 2)

Copyright

© 2021, Lim 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.

Metrics

  • 415
    Page views
  • 74
    Downloads
  • 0
    Citations

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)

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

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

Further reading

    1. Developmental Biology
    2. Neuroscience
    David Sokolov et al.
    Research Article

    Despite mounting evidence that the mammalian retina is exceptionally reliant on proper NAD+ homeostasis for health and function, the specific roles of subcellular NAD+ pools in retinal development, maintenance, and disease remain obscure. Here, we show that deletion of the nuclear-localized NAD+ synthase nicotinamide mononucleotide adenylyltransferase-1 (NMNAT1) in the developing murine retina causes early and severe degeneration of photoreceptors and select inner retinal neurons via multiple distinct cell death pathways. This severe phenotype is associated with disruptions to retinal central carbon metabolism, purine nucleotide synthesis, and amino acid pathways. Furthermore, transcriptomic and immunostaining approaches reveal dysregulation of a collection of photoreceptor and synapse-specific genes in NMNAT1 knockout retinas prior to detectable morphological or metabolic alterations. Collectively, our study reveals previously unrecognized complexity in NMNAT1-associated retinal degeneration and suggests a yet-undescribed role for NMNAT1 in gene regulation during photoreceptor terminal differentiation.

    1. Developmental Biology
    2. Neuroscience
    Nishtha Ranawat, Ichiro Masai
    Research Article

    Microglia are brain-resident macrophages that function as the first line of defense in brain. Embryonic microglial precursors originate in peripheral mesoderm and migrate into the brain during development. However, the mechanism by which they colonize the brain is incompletely understood. The retina is one of the first brain regions to accommodate microglia. In zebrafish, embryonic microglial precursors use intraocular hyaloid blood vessels as a pathway to migrate into the optic cup via the choroid fissure. Once retinal progenitor cells exit the cell cycle, microglial precursors associated with hyaloid blood vessels start to infiltrate the retina preferentially through neurogenic regions, suggesting that colonization of retinal tissue depends upon the neurogenic state. Along with blood vessels and retinal neurogenesis, IL34 also participates in microglial precursor colonization of the retina. Altogether, CSF receptor signaling, blood vessels, and neuronal differentiation function as cues to create an essential path for microglial migration into developing retina.