|Home | About | Journals | Submit | Contact Us | Français|
To gain insight into the mechanisms of Lmx1b function during ocular morphogenesis, we have studied the roles of lmx1b. 1 and lmx1b.2 during zebrafish eye development. In situ hybridization and characterization of transgenic lines in which GFP is expressed under lmx1b.1 regulatory sequence show that these genes are expressed in periocular tissues and in a pattern conserved with other vertebrates. Anti-sense morpholinos against lmx1b.1 and lmx1b. 2 result in defective migration of periocular mesenchymal cells around the eye and lead to apoptosis of these cells. These defects in the periocular mesenchyme are correlated with a failure in fusion of the choroid fissure or in some instances, more severe ventral optic cup morphogenesis phenotypes. Indeed, by blocking the death of the periocular mesenchyme in Lmx1b morphants, optic vesicle morphogenesis is largely restored. Within the retina of lmx1b morphants, Fgf activity is transiently up-regulated and these morphants show defective naso-temporal patterning. Epistasis experiments indicate that the increase in Fgf activity is partially responsible for the ocular anomalies caused by loss of Lmx1b function. Overall, we propose zebrafish lmx1b.1 and lmx1b. 2 promote the survival of periocular mesenchymal cells that influence multiple signaling events required for proper ocular development.
During eye development, the optic vesicles evaginate from the forebrain neuroepithelium and come into contact with several other tissues that are required for correct ocular morphogenesis, patterning and differentiation. For instance, contact with the prospective lens epithelium is correlated with invagination of the optic vesicle to form the twolayered optic cup consisting of prospective neural retina and pigment epithelium. Several recent studies have suggested that periocular mesenchyme derived from head mesoderm and neural crest cells also plays roles in the early morphogenesis of the optic cup and differentiation of the retina (Fuhrmann et al. 2000; Matt et al., 2005; Lupo et al., 2005; Molotkov et al., 2006; Matt et al. 2008).
Periocular mesenchyme eventually gives rise to specialized structures of the anterior segment of the eye that are responsible for aqueous humor dynamics (Johnston et al., 1979; Trainor & Tam, 1995; Cvekl and Tamm, 2004; Gage et al., 2005). Consequently, in humans disruption to factors important for the survival, migration and/or differentiation of periocular mesenchymal cells can lead to dysgenesis of the ocular anterior segment, elevated intraocular pressure and increased risk of glaucoma (Gould et al., 2004).
The glaucomas are a group of complex diseases characterized by vision loss due to damage of the optic nerve. Progress has been made towards identification of a subset of the critical risk genes for glaucoma. Several of these are also essential for ocular anterior segment development. Examples include genes encoding the transcription factors PITX2, FOXC1, and LMX1B. Mutations in either PITX2 or FOXC1 result in Axenfield-Reiger syndrome, in which a subset of affected individuals develop glaucoma (Walter 2003). The roles of PITX2 and FOXC1 in eye development have been explored in several species (Nishimura et al., 1998; Mears et al., 1998; Kidson et al., 1999; Smith et al. 2000; Semina et al., 1996; Gage et al., 1999; Tamimi et al., 2006). However, the role of LMX1B during ocular morphogenesis is less well understood.
LMX1B encodes a LIM-homeodomain transcription factor that when mutated in humans causes Nail-Patella Syndrome (NPS), a pleiotropic condition where approximately 50% of patients develop elevated IOP and glaucoma (Dreyer et al., 1998; Vollrath et al., 1998). In addition to the eye, NPS affects joint and limb development, and often disrupts function of the renal and central nervous systems (CNS). Genetic and biochemical analyses have shown that NPS is due to LMX1B haploinsufficiency (Lichter et al., 1997; Vollrath et al., 1998). Unlike NPS patients that are heterozygous for LMX1B mutations, mice heterozygous for null mutations appear normal. However, homozygous mutant mice display several abnormalities observed in NPS patients including abnormal development of renal structures, CNS patterning defects and dysgenesis of anterior ocular tissues (Chen et al., 1998; Kania et al., 2000; Guo et al., 2007; Pressman et al., 2000). However, it remains unknown how LMX1B regulates ocular development.
To better understand the mechanisms by which Lmx1b defects cause ocular pathology, we examined the expression and loss-of-function phenotypes of lmx1b. 1 and lmx1b.2 in zebrafish. Analysis revealed that both genes have expression patterns conserved with other vertebrates, including within the periocular mesenchyme. Loss-of-function analyses showed Lmx1b activity is required for periocular mesenchymal cell survival, optic cup morphogenesis and choroid fissure closure. In addition, knock-down of Lmx1b in transgenic lines that express GFP under lmx1b. 1 regulatory sequence revealed migration defects in periocular mesenchyme. We also found that lmx1b morphants showed increased ocular FGF activity, and consistent with this, abnormal naso-temporal patterning of the retina (Picker and Brand, 2005). Preventing apoptosis of periocular mesenchyme in lmx1b morphants alleviated the morphogenesis defects. Furthermore, reducing Fgf activity partially restored retinal patterning. Together, these results suggest that altered Fgf signaling due to periocular mesenchymal cell death is a contributing factor to the ocular pathogeneses associated with loss of Lmx1b activity.
Zebrafish embryos were raised at 28.5°C and staged according to Kimmel et al. (1995). Phenylthiourea (PTU) was applied to embryos to prevent melanization when necessary.
Total RNA was extracted from adult zebrafish eyes using RNAeasy Mini Kit (Qiagen). RT-PCR combined with 5’ and 3’ RACE was performed to identify full-lenfth lmx1b.1. To isolate lmx1b.1 full-length sequence, the following primers were used:
Full-length lmx1b.2 sequence was isolated by RT-PCR using the following primers:
Zebrafish lmxb.1 (AY551077) and lmx1b.2 (AY551078) sequences have been deposited into GenBank.
Embryos were dechorionated and fixed overnight at 4°C in 2.5% gluteraldehyde/1 % paraformaldehyde in phosphate buffered sucrose, pH7.4. The next morning embryos were dehydrated and infused with Epon. Transverse sections were 1 µm thick, heat-mounted on gelatin coated glass slides, and stained with 1 % toluidine blue.
Whole mount in situ hybridization was performed as previously described (Thisse and Thisse, 2004) with one modification. Following LiCl precipitation, antisense RNA probe was further purified using a ProbeQuant G-50 micro spin column (GE Heathcare).
Morpholino oligonucleotides (GeneTools, Inc.) were targeted to splice site junctions between exon 1 and intron 1 (splice-MOs) and the translation start site (ATG-MOs) for each lmx1b.1 or lmx1b.2. The following morpholino sequences were used:
Mismatch controls had sequences matching those of the splice-inhibiting morpholinos with the exception of 5 bases denoted in lowercase letters:
For each lmx1b gene, translation-inhibiting and pre-mRNA splice-inhibiting morpholinos were injected at the one-cell stage with 1% phenol red (Nasevicius and Ekker, 2000). For lmx1b.1 and lmx1b.2, both translation- and splice-inhibiting morpholinos produced similar phenotypes. As controls, morpholinos with 5 base-pairs mismatched relative to the splice-disrupting oligos did not produce phenotypes when injected at similar concentrations to those used for the experimental morpholinos. Each splice disrupting morpholino inhibited normal transcript processing while mismatch control morpholinos did not. Quantitative real-time RT-PCR determined that the lmx1b splice-inhibiting morpholinos functioned as predicted. For lmx1b morphant experiments described in the results, splice-inhibiting morpholinos were used and for most analyses, embryos showing moderate phenotypes were analyzed (see Fig. 2 and Table S1 for details).
RT-PCR was used to isolate −5kb of the lmx1b. 1 promoter using the following primers designed from HTG sequence clone CH211-81G2:
5’ GGGGACAACTTTGTATAGAAAAGTTGATCACGTGCTTTTGGGTTTC and Lmx1b. l-5kbR
5’ GGGGACTGCTTTTTTGTACAAACTTGGCGGATGATCTTCGATTTT. Gateway technology (Invitrogen) using Tol2-kit reagents (Kwan et al. 2007) was employed to generate tol2:-5lmx1b.1:lmx1b:ires:GFP and tol2:-5lmx1b.1:GFP. Expression of the tol2 constructs was achieved by co-injection of transposase mRNA as previously described (Kawakami 2005).
Two independent sets each of 18 somite stage and 24 hours post fertilization (hpf) embryos, each analyzed in triplicate, injected with either lmx1b.1, lmx1b.2, or lmx1b.1+lmx1b.2 (lmx1bdMO) splice-inhibiting morpholinos were collected for total RNA isolation using the QIAGEN RNeasy Plus Mini Kit (QIAGEN Inc.). Age-matched uninjected embryos were collected and RNA was isolated in parallel with morphant RNA. The degree of gene knockdown was evaluated as the average fold-reduction as compared to age-matched uninjected embryos. cDNA was produced using 0.5 µg of RNA with the SuperScript II Reverse Transcriptase enzyme (Invitrogen). qPCR reactions were carried out using the iQ SYBR Green Supermix (Bio-Rad) on the iCycler iQ Real-Time Detection System (Bio-Rad). All cDNA samples were assayed in triplicate for each primer set. Primers to amplify a 240 base pair (bp) fragment of lmx1b.1 were designed (F: 5’-TTGGACGGTATAAAAATCGAAGA-3’; R: 5’-TTTGCAGTAAAGTTTCCTCTCC-3’)- Primers to amplify a 249 bp fragment of lmx1b.2 were designed (F: 5’-ATGCTGGACGGAATCAAAATT-3’; R: 5’-AGTCGTGTTTACAGTAGAGTTTGTGG-3’)- Amplification of ef1 alpha was used for normalization (F: 5’-TGGGCACTCTACTTAAGGAC-3’; R: 5’-TGTGCCAACAGGTGCAGTTC-3’).
TUNEL labeling to detect apoptosis was performed using the Apoptag kit (Chemicon International). Manufacturers instructions were followed for the labeling reaction and embryos were then washed in phosphate-buffered saline, blocked and developed as described for a standard whole mount in situ hybridization/antibody labeling protocol (Shanmugalingam et al. 2000).
Antibodies to activated Caspase-3 (affinity purified rabbit polyclonal, R &D Systems Cat #AF835, Lot#CFZ32) and phospho(ser10)Histone-H3 (rabbit polyclonal, Upstate – Millipore Cat#06–570, Lot#28770) were both used on 4% paraformaldehyde-fixed cryosections at 1:1000 dilution.
The Ffg-receptor inhibitor SU5402 (Calbiochem) was stored as 10mM stock solutions in DMSO and added to 6 somite stage embryos at a final concentration of 5 µM. Control embryos were treated with DMSO at equivalent concentrations.
Blastulae transplantation was performed as previously described to generate chimeric embryos (Ho and Kane, 1990). Embryos derived from matings of Tg(foxd3:GFP)zf15 and Tg(h2afx:H2A-mCherry)mw3 were used as donors. Approximately equivalent numbers of cells (~20–25 cells) were transplanted in wild-type unlabeled hosts after the 1000 cell stage and before the dome-stage. Following the transplantation of blastulae cells into the ventral blastoderm margin, a region fated for neural crest, mosaic embryos were raised to 30 hpf. At 30 hpf, mosaic embryos were screened for at least 1 mCherry-positive nuclei in the ocular mesenchyme to indicate that neural crest targeting was achieved, and these specimens were fixed in 4% paraformaldehyde for quantitative analysis. Wild-type and lmx1bdMO donor mosaics were scored for the number of H2A-mCherry and foxD3:GFP-positive cells using a compound fluorescent microscope. Two separate experiments were carried out to achieve an n=12 embryos for each condition. A third experiment was carried out to compare mismatch lmx1bdMO donor cell behavior to that of uninjected wild-type donor cells. No differences were noted between these two control donor cell types.
Transcripts for lmx1b.1 and lmx1b.2 are expressed dynamically and largely in overlapping domains during embryonic development (Fig. 1). Before the 13 somite stage (s), both genes were expressed in the prospective midbrain-hindbrain boundary (MHB) and axial midline tissue (O’Hara et al., 2005). At 18 s, both lmx1b.1 and lmx1b.2 were expressed in the MHB, ventral diencephalon, and midline tissues (Fig. 1A,G) and at 24 and 36 hpf, in the ventral diencephalon, midline, fin buds, nephric primordia, MHB and hindbrain cells (Fig. 1B–E; H–K). We observed unique expression of lmx1b.1 in the dorsal diencephalon and otic vesicles at 18 s (Fig. 1A), and in spinal cord cells (Fig. 1B,E boxed). By 42 hpf, lmx1b.1 expression was maintained in diencephalic, midbrain and hindbrain cells, and fins (Fig. 1F). lmx1b.2 transcripts were detected in domains overlapping with lmx1b.1, but expression was more restricted (Fig. 1H–L). Indeed, the expression domains of lmx1b.1 were broader than those of lmx1b.2 throughout all developmental stages analyzed.
Within the ocular region, transcripts for both lmx1b genes were detected in periocular cells associated with the optic stalk (Fig. 1C,I) and periocular mesenchyme underlying the surface ectoderm at 24 hpf (Fig. 1D,J). By 36 hpf, lmx1b.1 expression persisted in periocular mesenchyme and was enriched in presumptive anterior segment structures and cells of the optic fissure (Fig. 1E, K). By 42 hpf, lmx1b.1 expression was also associated with the hyaloid vasculature (data not shown). Similar to lmx1b.1, lmx1b.2 was expressed in periocular mesenchyme at 24 hpf, and in cells lining the optic fissure by 36 hpf (Fig. 1I–K).
Overall, the expression patterns of the duplicated lmx1b genes were overlapping and highly conserved when compared with those of higher vertebrate orthologues (Dunston et al., 2005). Overlapping expression patterns suggest at least partial redundancy of function and previous studies have indeed shown that lmx1b.1 and lmx1b.2 have redundant functions in maintaining the isthmic organizer (O’Hara et al., 2005).
To explore the role of lmx1b.1 and lmx1b.2 in eye development, anti-sense morpholinos were used to diminish their function. For each lmx1b gene, translation-inhibiting and pre-mRNA splice-inhibiting morpholinos were tested. Both types of morpholinos produced CNS phenotypes as previously observed (O’Hara et al., 2005; Filippi et al., 2008); however, the splicing morpholinos were more efficient and were used for our analyses (see Materials and methods, Supplemental Fig.S1, and Table S1 for details).
By 24 hpf, embryos injected with either lmx1b.1 or lmx1b.2 morpholinos showed defects in ventral eye morphogenesis. Co-injection of lmx1b.1 and lmx1b.2 morpholinos, herein referred to as lmx1bdMO morphants, typically resulted in more severe ocular phenotypes as compared to single morphants, but still yielded a spectrum from mild to severe (Fig. 2A–H; Supplemental Fig. S1; Table S1). In the mild phenotype, early morphogenesis of the eye cup was overtly normal, but by 36 hpf the choroid fissure failed to close (lines in Fig. 2F), a condition called coloboma. This ventral ocular defect was more evident in embryos showing moderate and severe phenotypes. In these embryos the eye lost its spherical shape and was often completely open on the ventral side (Fig. 2G,H). In addition, morphants eyes often appeared smaller with increased pyknotic cells as compared to controls (Fig. 2E–H). With regard to the variability in phenotypes, the efficiency of normal transcript splicing was disrupted in a manner that correlated with mild versus severe classes. However, we also found that the proportion of mild versus severe phenotypes varied with the genetic background of the injected embryos. Together this suggests that Lmx1b function is both dosage-sensitive and can be modified by genetic background.
Histological analysis of 24 hpf lmx1bdMO morphant eyes exhibiting a moderate phenotype showed a thinned retinal neuroepithelium and loosely organized periocular mesenchyme, with increased numbers of pyknotic nuclei, particularly in the ventral region (Fig. 2Q,R). At 62 hpf, when retinal lamination was apparent in control eyes, laminar differentiation was delayed and ventral ocular tissue often appeared disorganized in lmx1bdMO morphants (Fig. 2S,T). In these embryos, the retinal pigment epithelium often appeared reduced and disorganized in the ventral retina (Fig. 2T)
Because lmx1b genes are expressed in periocular mesenchyme cells, we next assessed for changes in expression of markers of this population of cells that may accompany the eye defects in the lmx1bdMO-injected embryos. In mild and moderate lmx1bdMO phenotypes, we observed that foxc1a and eya2 expressing periocular mesenchyme cells were able to migrate to the eye despite the loss of Lmx1b activity. In some cases, we detected an accumulation of periocular mesenchyme at the choroid fissure (Fig. 2J,N) and in more affected embryos, periocular cell migration towards the nasal-ventral region of the eye was altered (asterisk in Fig. 2O; Fig 7). This suggests a defect in migration of periocular mesenchymal cells to defined locations of the eye. In severely affected lmx1bdMO morphants, periocular expression of foxc1a and eya2 was almost absent, suggesting either a loss of marker gene expression or failure of periocular mesenchyme cells to arrive to the eye (Fig. 2L,P). Consistent with disruptions to periocular mesenchyme, vascular patterning was affected in the eyes of lmx1bdMO morphants at later stages (Supplemental Fig. S2). This data suggests that the most severe ocular phenotypes could be due to a loss of periocular mesenchyme, but that the mild and moderate ocular phenotypes cannot simply be due to the absence of these cells.
Knockout of lmx1b in mice leads to altered regulation of differentiation markers in anterior segment cells originating from periocular mesenchyme (Pressman et al., 2000). Histological inspection of lmx1b morphants suggested that in zebrafish loss of lmx1b in zebrafish results in apoptosis as well as differentiation defects in periocular mesenchyme. To assess apoptosis in lmx1bdMO morphants, we performed TUNEL analysis which revealed significantly elevated cell death in regions where lmx1b genes are normally expressed, most notably at the MHB and in cells around the eye (Fig. 3A, middle). Depletion of p53, a protein required for apoptosis under many conditions, blocked the regionally-localized cell death observed in lmx1bdMO embryos (Fig. 3A, right). Furthermore, overall eye size defects and ventral ocular dysgenesis were abrogated when apoptosis was inhibited, as observed using transgenic embryos in which individual cells were highlighted by expression of beta-actin-GFP (Tg(Bactin:HRAS-EGFP)vu119; Cooper et al., 2005) (Fig 3B).
When apoptotic cells were labeled in transgenic morphant embryos expressing GFP in neural crest cells (Tg(−7.2sox10:EGFP)zf77; Hoffman et al., 2007) there was minimal co-localization at 36 hpf (Fig. 3C). However, neural crest cells were depleted from the ventral eye and apoptotic cells were numerous in this region (Fig. 3C). This suggests that the GFP in the neural crest cells was degraded by the time that the cells became TUNEL-positive. Additionally, some apoptotic periocular cells may be mesoderm-derived and therefore did not express the transgene.
To more directly evaluate whether cell death in embryos injected with lmx1b morpholinos was due to specific loss of Lmx1b function and not to off-target effects, we first attempted to rescue apoptosis by co-injecting lmx1b mRNA with morpholinos. However, global expression of either lmx1b.1 or lmx1b.2 mRNA resulted in severe gastrulation defects. To more faithfully express lmx1b cDNAs in endogenous locations, we isolated 5 kilobases (kb) of sequence directly upstream of the lmx1b. 1 gene and utilized this sequence to drive GFP expression (tol2:-5lmx1b:GFP). This tol2-based construct resulted in GFP accumulation in regions similar to that of endogenous lmx1b. 1 mRNA expression (Fig. 4A). We also generated a tol2-based expression construct where the 5kb lmx1b. 1 promoter sequence was placed in front of the lmx1b. 1 cDNA, followed by an internal ribosome entry site (ires) proximal to GFP (tol2:- 5lmx1b.1:lmx1b:ires:GFP). This construct was co- injected with lmx1b morpholinos into transgenic, H2A-mCherry embryos (Tg(h2afx:H2A-mCherry)mw3 ). Expression of the nuclear-localized mCherry fluorescent protein revealed condensed chromatin of pyknotic cells when imaged at sub-threshold gain.
Expression of lmx1b. 1 cDNA under the control of its endogenous −5kb promoter rescued much of the cell death associated with injection of the lmx1bdMO (Fig. 4B,D). Expression of GFP alone did not prevent morpholino-induced cell death (Fig. 4B,C). Together these data suggests that loss of Lmx1b activity results in apoptosis of periocular mesenchymal cells, which normally regulate ocular patterning and morphogenesis.
To better evaluate the consequences of Lmx1b depletion on lmx1b-positive cells, we established four independent transgenic lines using the tol2:-5lmx1b.GFP construct (Tg(-5kblmx1b.1:GFP)mw10–13). Each line showed a high degree of specific overlap between GFP expression and endogenous lmx1b. 1 expression. Using cryosections from Tg(-5kblmx1b.1:GFP)mw11 embryos, we further assessed the pattern of cell death and proliferation for Imx1 b. 1:GFP-positive cells following injection of control or lmx1bdMO oligonucleotides. Immunoreactivity of activated Caspase-3, a marker of apoptotic cells, confirmed TUNEL analyses and showed that both lmx1b.1: GFP-positive and lmx1b.1: GFP-negative periocular cells displayed elevated apoptosis following depletion of Lmx1b activity (Fig 5A,B). For phosphoHistone-3 immunoreactivity, which marks cells in late G2/M-phase of the cell cycle, we did not find differences between control and lmx1bdMO eyes (Fig 5C,D). However, proliferation of periocular mesenchymal cells in either control or lmx1bdMO eyes was rare.
Periocular cell death and differentiation defects in lmx1bdMO morphants may still be due to secondary defects caused by loss of Lmx1b function in the ventral diencephalon. To address this possibility, we generated genetic mosaic embryos in which small groups of cells with compromised Lmx1b activity were located in otherwise wild-type embryos. Donor embryos were derived by mating Tg(foxd3:GFP)zf15 and Tg(h2afx:H2A-mCherry)mw3 transgenic fish in order to label all donor cells with the H2A-mCherry transgene and periocular neural crest donor cells with the foxD3:GFP transgene (Fig. 6A). Hosts embryos were always derived from non-transgenic, wild-type parents. Comparisons were made between lmx1bdMO-injected and non-injected donor cells to address the autonomy of Lmx1b activity on periocular mesenchymal cell apoptosis and migration defects. Following the transplantation of blastulae cells into regions fated for neural crest, mosaic embryos were fixed at 30 hpf. To ensure that neural crest targeting was achieved, embryos were screened for at least one mCherry-positive nucleus in the ocular mesenchyme and none in the ventral forebrain. Wild-type and lmx1bdMO donor mosaic embryos were scored for the number of H2A-mCherry and foxD3:GFP positive periocular cells. Host embryos with donor cells deficient for Lmx1b function showed significant reduction in both the total number of donor cells in the periocular region and foxD3:GFP-positive cells (Fig. 6B,C). Like embryos in which lmx1b genes were knocked-down globally, lmx1b-deficient periocular mesenchymal cells tended to accumulate dorsally and posteriorly within wild-type host eyes (Fig. 6B, lower right). In addition to periocular mesenchyme, mosaic embryos from either condition showed equivalent contributions to non-ocular structures, indicating that wild-type and lmx1bdMO donor cells transplanted with similar efficiencies. Furthermore, mismatch control morpholino donor cells showed similar distribution and numbers of periocular mesenchymal cells as compared to wild-type donor cells.
These results indicate that the lmx1b genes function cell-autonomously in periocular cells to regulate their survival and potentially normal migration. Evidence supporting a role for periocular mesenchyme in eye patterning has been previously reported (Fuhrmann et al. 2000; Matt et al., 2005; Lupo et al., 2005; Molotkov et al., 2006; Matt et al. 2008). For these reasons, we explored the consequences of abrogation of Lmx1b function first on periocular cell migration and then on retinal morphogenesis and patterning.
To directly evaluate lmx1b-positive periocular cell migration we injected Tg(−5kb lmx1b.1:GFP)mwl0 and Tg(−5kblmx1b.1:GFP)mw11 embryos with equal concentrations of either mismatch control morpholinos or the lmx1bdMO cocktail. Between 20–60 hpf, the location of GFP-positive cells was assessed in control or lmx1bdMO embryos. We found that lmx1b.1:GFP expression was significantly reduced within the optic stalk region throughout the period that this transgene was expressed there (21–30 hpf) (Fig. 7A,D arrows; Supplemental Fig. S3). When periocular cells arrived to the eye, the number of GFP-positive cells were equivalent in control and lmx1bdMO embryos (Fig. 7A,D). By 36 hpf, the number of lmx1b. 1 :GFP-positive cells was reduced and their location was altered. Specifically, in control embryos GFP-positive cells began to accumulate in a more anterior/nasal region, but GFP-positive cells tended to remain in a more posterior/temporal location in lmx1bdMO eyes (Fig. 7B,E). These differences were maintained throughout the efficacy of morpholino knock-down (Fig. 7C,F). This experiment, along with the marker studies described above, suggests that loss of Lmx1b results in both elevated apoptosis and migration defects for periocular mesenchymal cells.
The ventral ocular defects and coloboma phenotype observed in lmx1bdMO morphants are reminiscent of defects caused by alteration to pax2, vax1 and vax2 expression in the optic stalk and ventral neural retina (Macdonald et al., 1996; Take-uchi et al., 2003). Therefore we analyzed the expression pattern of these transcripts in lmx1bdMO morphants exhibiting mild to moderate eye morphogenesis phenotypes.
As compared to 24 hpf uninjected controls, we observed expansion of pax2.1 expression in the optic stalk and nasal retina of lmx1bdMO morphants (Fig. 8A,B). At 24 hpf, vax1 and vax2 are normally expressed in the optic stalk and preoptic area (Fig. 8C,E) and vax2 is additionally expressed in the ventral retina. As compared to control embryos at 24 hpf, vax1 expression was expanded in the ventral region of the optic cup in lmx1bdMO morphants, including cells within the presumptive retinal pigmented epithelium (Fig. 8C,D). Expression of vax2 was also enhanced in the optic stalk and ventral retina, and slightly expanded into more nasal and temporal regions of the ventral retina (Fig. 8F). Because lmx1b genes are not obviously expressed directly within the ventral retina, these results indicate that Lmx1b activity in non-retinal cells, most probably periocular mesenchymal cells or those associated with the optic stalk, are required for ventral retinal development.
Fgfs are implicated in several aspects of ocular development, including optic cup patterning, retinogenesis, and anterior segment morphogenesis. For instance, Fgf activity from the optic stalk contributes to vax regulation (Takeuchi et al., 2003) and when Fgf signaling is increased in the optic stalk and periocular regions by the aussicht mutation, defects in optic fissure closure, ocular patterning and differentiation are observed (Heisenberg et al., 1999). To determine if ocular dysgenesis in lmx1b morphants involves altered Fgf signaling, we analyzed fgf3 and fgf8a expression patterns in lmx1b morphants. At 18 s,fgf3 expression in lmx1bdMO morphants appeared normal (data not shown). By 24 hpf, lmx1bdMO morphants revealed a slight alteration of fgf3 expression in the rostral CNS and complete down-regulation in the MHB (Fig. 9A–D and O’Hara et al., 2005). Morphants displayed a more dramatic change in the pattern of fgf8a. At 18 s, fgf8a expression in the optic stalk region was moderately up-regulated in lmx1bdMO morphants (data not shown) and by 24 hpf, significantly expanded within the optic stalk, ventral retina and telencephalon (Fig. 9F–H).
In reciprocal experiments, we performed expression analysis of lmx1b genes in fgf3, fgf8a, and fgf3+fgf8a morphants. At 24 hpf, forebrain and ocular expression of lmx1b. 1 and lmx1b.2 was not significantly affected in the fgf3, fgf8a or fgf3+fgf8a morphants (Supplemental Fig. S4). However, subtle changes were observed within the MHB, perhaps owing to the essential role of Fgf3 and Fgf8a on the formation of this structure.
To address whether Fgf-mediated gene transcription was affected by compromising Lmx1b function, we analyzed various target genes of Fgf signaling in ocular regions including erm, pea3, spry2, and dusp6:GFP (Roehl and Nusslein-Volhard, 2001; Furthauer et al., 2002; Tsang et al., 2002; Raible and Brand, 2001; Molina et al., 2007). Upon knock-down of both lmx1b genes, all Fgf-regulated genes assessed were significantly up-regulated in forebrain and eye tissues, but down-regulated in the MHB region (Fig. 10). Cryosections of dusp6:GFP, as well as adjusted-focus inspection of whole-mount stained embryos, indicated the Fgf target genes were up-regulated in the neural retina, optic stalk regions, and lens, but not in periocular cells. Together, these results suggest that Lmx1b negatively regulates ocular fgf3 and fgf8a expression and activity.
Fgf signals, including Fgf8, determine nasal-temporal patterning of the retina by promoting nasal and repressing temporal identity (Picker and Brand, 2005). Given that lmx1bdMO morphants showed an up-regulation of Fgf signaling in the eye, we assessed whether loss of Lmx1b function affected nasal-temporal patterning of the retina. Indeed, the nasal marker efna5a and dorsal marker tbx5 were both expanded in lmx1bdMO morphants (Fig. 11A,C). Conversely, the temporal marker epha4b was reduced following loss of lmx1b genes (Fig. 11B).
In order to evaluate the relationships between Lmx1b and Fgf signaling, we analyzed the consequences of loss of lmx1b function in the fgf8a mutant acerebellar (ace). As already reported, the expression of tbx5 and efna5a were reduced and epha4b expanded in fgf8a/ace mutants (Fig 11; Picker and Brand 2005). Loss of Fgf8a activity in lmx1bdMO morphants led to phenotypes intermediate between control and Lmx1b-deficient conditions. This suggests that some of the consequences of loss of Lmx1b function are dependent upon Fgf8 signaling.
To more severely abrogate Fgf signaling in lmx1b morphants, Fgf receptor activity was inhibited with SU5402 starting at the 6 somite stage. In agreement with a previous study (Picker and Brand, 2005), SU5402-treated embryos showed more severe naso-temporal patterning defects (Fig.11A–C). Although SU5402 treatment of lmx1bdMO morphants partially restored nasal and temporal expression patterns, tbx5 remained expanded in the dorsal retina and embryos still showed ventral ocular dysgenesis (Fig. 11B). Together, these experiments suggest that increased Fgf activity in lmx1bdMO embryos is partially, but not fully, responsible for the ocular defects caused by loss of Lmx1b function.
In this study, we describe the expression and consequences of loss of function of lmx1b. 1 and lmx1b. 2 during zebrafish ocular morphogenesis. As previously reported for other vertebrates, lmx1b.1 and lmx1b.2 are expressed in periocular mesenchymal cells. Here we show that together they are required cell-autonomously for migration and survival of these cells. Lmx1b has a cell-non-autonomous requirement for ventral eye morphogenesis, closure of the choroid fissure and retinal pattering. Finally, levels of Fgf signaling are dependent upon Lmx1b activity and enhanced Fgf activity is partially responsible for retinal patterning phenotypes observed in Lmx1b morphants.
Early in development, the emerging eye is surrounded by periocular mesenchyme that contributes to specialized non-neural structures of the eye after morphogenesis is complete. Emerging evidence suggests that reciprocal interactions between the periocular mesenchyme and the retina are required for the correct morphogenesis of the eye. The retina plays a critical role in directing the migration of the periocular mesenchyme and once these cells are in place, signals such as retinoic acid (RA) act in a paracrine manner to regulate expression of the periocular mesenchyme genes such as pitx2, eya2 and foxc1 (Matt et al., 2005; Molotkov et al., 2006; Langenberg et al., 2008; Matt et al., 2008). Here we suggest that reciprocally, the periocular mesenchyme is necessary for the proper morphogenesis of the ventral eye. In addition, it is possible that Imx 1b-positive cells of the ventral forebrain also influence ocular patterning. However, unlike lmx 1b-positive periocular cells, the ventral forebrain cells are not overtly affected by loss of Lmx1b function. Indeed, the primary cause for ocular dysgenesis and coloboma upon loss of Lmx1b function appears to be apoptosis of periocular mesenchyme associated with the optic stalk and globe of the eye, as the visible ocular phenotypes following knock-down of lmx1b are rescued by blocking apoptosis. A similar requirement for signaling from the periocular mesenchyme has been proposed in mice, as impairing RA signaling from neural crest cells is sufficient to alter eye morphogenesis (Molotkov et al. 2006; Matt et al., 2008). Disrupting RA signaling appears to affect morphogenesis but not axial patterning, whereas upon Lmx1b knockdown, we observed significant defects in both dorso-ventral and naso-temporal patterning. This suggests that other signaling pathways, such as Fgf, are likely involved in the reciprocal cross-talk between periocular mesenchyme and retina. Our studies, combined with those of others, point to the requirement for complex interactions between periocular mesenchyme and retina for patterning and morphogenesis. Notably, disrupting this cross talk gives rise to coloboma, a common yet poorly understood ocular defects.
Lmx1b functions in various embryonic structures, and loss-of-function phenotypes have been described in humans, mice, and zebrafish (Dryer et al., 1998; Vollrath et al., 1998; Chen et al., 1998; O’Hara et al., 2005). In this paper we report a new role for Lmx1b in ocular morphogenesis. In humans, heterozygous loss-of-function mutations result in Nail Patella Syndrome. These patients are prone to develop glaucoma, but do not show visible ocular dysmorphogenesis (Lichter et al., 1997). This phenotype is potentially the result of subtle developmental defects that sensitize individuals to glaucoma. It is difficult to directly compare the zebrafish with the human phenotypes given that homozygous LMX1B mutations have not been observed and likely cause early embryonic lethality. In contrast to humans, heterozygous loss-of-function mutations in mouse lmx1b do not result in obvious phenotypes, although extensive analysis on multiple genetic backgrounds has not been described. Homozygous null mutations in mice do result in ocular defects but they are considerably milder than those seen in zebrafish morphants (Pressman et al., 2000). In the mouse eye, lmx1b −/− mutations primarily affect differentiation, and not survival or migration of periocular mesenchyme. Furthermore the dorso-ventral differences in the severity of the fish eye phenotypes are not evident in mouse mutants. However, the normal dorso-ventral anatomical differences in the zebrafish anterior segment are much more striking than in mammals (Soules and Link, 2005). One similarity is that in both zebrafish morphants and mouse mutants, the overall eye size is often small, suggesting a conserved non-cell autonomous role for Lmx1b in ocular growth. Furthermore, for all vertebrates analyzed Lmx1b is required for cell survival within MHB tissues. None-the-less, while Lmx1b function is strongly conserved in its requirement for cell survival in isthmo-cerebellar region, Lmx1b has a more critical role in eye development for fish than mouse.
Although different thresholds for Lmx1b are required for humans and mice, lmx1b phenotypes have relevance to anterior segment development and glaucoma in both species. Similar to other species, we find zebrafish lmx1b is expressed in periocular mesenchyme cells that contribute to anterior segment structures. In addition, we observed a range of eye phenotypes of differing severity following loss of Lmx1b function, the milder of which may be more analogous to the human phenotypes and have relevance to glaucoma. In these zebrafish morphant phenotypes, the expression patterns of foxc1 (Fig. 2) and pitx2 (data not shown), key regulators of anterior eye segment development and risk genes for glaucoma, were disturbed. We observed accumulations of both lmx1b and foxc1 -expressing cells associated with defects in their migration towards the nasal-ventral region of the eye and into the eye-cup. Accumulations of foxc1- and eya2-expressing cells in the ventral part of the eye are unlikely due to increased apoptosis and are more likely a consequence of abnormal migratory behavior or defects in differentiation. Alterations in the periocular location of lmx1b.1:GFP-positive cells at multiple time-points more directly demonstrates defects in cell migration. Given the limitations of long-term gene knock-down using morpholinos, we did not determine the late consequences of Lmx1b depletion on periocular mesenchyme differentiation. However, it is likely that prolonged lmx1b knock-down would lead to anterior segment dysgenesis. Thus future studies are needed to determine if permanent reduction of Lmx1b function leads to anterior segment dysgenesis and other glaucoma-associated phenotypes in zebrafish.
Lmx1b is essential for transcriptional network regulation during development in multiple organ systems. For example, Lmx1b specifies dorsal-ventral identity in the limb bud mesenchyme (Chen and Johnson, 2002) and as part of its organizing function in this tissue, Lmx1b non-autonomously regulates motorneuron trajectory (Kania et al., 2000). Lmx1b is also essential for the appropriate regulation of organizing activity in the MHB (O’Hara et al., 2005; Matsunaga et al., 2002; Guo et al., 2007; this study). In this region, Lmx1b maintains other factors, such as pax2, that promote fgf8 transcription (O’Hara et al., 2005). In agreement with that study, we found diminished fgf8 expression in the MHB of lmx1bdMO embryos. In contrast to the MHB, within the eyes loss of Lmx1b function resulted in up-regulation of Fgf signaling.
Fgf signaling promotes nasal and represses temporal retinal identity (Picker et al. 2005) and we observed similar results from Lmx1b loss of function experiments. In our experiments, Fgf and Lmx1b act antagonistically and reduction of both factors are closer to wild-type than phenotypes resulting from loss of either Fgf signaling or Lmx1b function alone. However, epistasis cannot solely explain the antagonistic relationship because blocking Fgf signaling only partially rescued the loss of Lmx1b function. Thus, loss of Lmx1b likely affects additional signaling pathways to cause ocular patterning and ventral eye morphogenesis defects. As Fgf functions together with RA and Hedgehog signaling during dorso-ventral patterning of the eye, these same signaling pathways are candidates for Lmx1b-mediated signaling (Lupo et al., 2005). Cumulatively, our studies combined with others demonstrate critical, but context-dependent functions for Lmx1b in coordinating signaling activities, including those mediated by Fgfs, to regulate cell proliferation, migration, fate-determination, and survival within multiple tissues.
This study was supported by grants F32EY16321 (CM), T32EY014537 (CM), R01EY014167 (BL), a Telethon Fellowship (GG), and grants from the MRC and Wellcome Trust (SW). We thank members of our labs for discussions and Pat Cliff, Melissa Reske, and Anitha Ponnuswami for technical help.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.