Journal of Molecular Medicine

, Volume 84, Issue 7, pp 604–615

Molecular dissection reveals decreased activity and not dominant negative effect in human OTX2 mutants


  • Gilles Chatelain
    • LBMC, ENS-Lyon, IFR128 Lyon-Gerland
  • Nicolas Fossat
    • LBMC, ENS-Lyon, IFR128 Lyon-Gerland
  • Gilbert Brun
    • LBMC, ENS-Lyon, IFR128 Lyon-Gerland
    • LBMC, ENS-Lyon, IFR128 Lyon-Gerland
Original Article

DOI: 10.1007/s00109-006-0048-2

Cite this article as:
Chatelain, G., Fossat, N., Brun, G. et al. J Mol Med (2006) 84: 604. doi:10.1007/s00109-006-0048-2


The paired-type homeodomain transcription factor Otx2 is essential for forebrain and eye development. Severe ocular malformations in humans have recently been associated with heterozygous OTX2 mutations. To document the molecular defects in human mutants, Otx2 structural characterization was carried out. A collection of deletion and point mutants was created to perform transactivation, DNA binding, and subcellular localization analyses. Transactivation was ascribed to both N- and C-termini of the protein, and DNA binding to the minimal homeodomain, where critical amino acid residues were identified. Acute nuclear localization appeared controlled by a nuclear localization sequence located within the homeodomain which acts in conjunction with a novel nuclear retention domain that we unraveled located in the central part of the protein. This region, which is poorly conserved among Otx proteins, was also endowed with dominant negative activity suggesting that it might confer unique properties to Otx2. Molecular diagnostic of human mutant OTX2 proteins discriminates hypomorphic and loss of function mutations from other mutations that may not be relevant to ocular pathology.


Otx2Ocular malformationTransactivationNuclear localizationDNA bindingDominant negative



Nuclear localization sequence


Amino acid


Bicoid target site


Together with Otx1 and Crx, the mammalian Otx2 transcription factor belongs to the Paired-class Otx homeoprotein family that is widely conserved among animal phyla [13]. Otx2 regulates key early developmental processes such as gastrulation movements and forebrain induction and maintenance [4, 5]. Later, it is involved in eye tissues formation and differentiation [68]. These multiple activities rely on spatially and temporally controlled expression via alternative promoters [9, 10] and remote complex enhancers [11, 12]. In some genetic backgrounds, Otx2+/− mice display cranio-facial abnormalities and microphtalmia [13].

Eight heterozygous OTX2 mutations were recently identified in patients suffering from severe ocular malformations [14]. Five of the eight coding-region sequence changes were predicted to cause premature protein truncation either through direct creation of a stop codon or through a frameshift, and three were missense mutations affecting conserved residues. To define the molecular defects in these mutants and decipher whether they act through decreased activity or dominant negative effect, we undertook a rationalized analysis of the function of Otx2 protein conserved domains. Using promoter activation, DNA binding, and subcellular localization assays, we delineated regions and residues critical for transactivation, DNA binding, and nuclear localization. We, thus, discriminated human mutations which lead to partial or complete haploinsufficiency and mutations that may be unrelated to the patients’ phenotype. In the course of this study, we uncovered a novel nuclear retention motif that can also exert dominant negative effects.

Experimental procedures


Otx2 full-length cDNA expression plasmid has been previously described [10]. A mouse-enhanced green fluorescent protein (GFP) expression plasmid pCMX-GFP [15] was a gift from F. Wianny. To make Otx2-GFP, the Otx2 TGA stop codon was converted to GGA, and a BamHI site was created by polymerase chain reaction (PCR)-directed mutagenesis. The GFP coding sequence was then amplified from pCMX-GFP using primers that introduced a 5′ BamHI site and a 3′ SpeI site and inserted in frame downstream the Otx2 sequence. Otx2 point mutants R89G, P133T, and P134A [14] were created by PCR-directed mutagenesis. To make all other GFP fusions, the GFP stop codon in pCMX-GFP was converted into a BamHI site allowing in-frame fusion to create pGFP-N. All full-length Otx2, point mutant, and truncated forms of Otx2 were amplified by PCR-utilizing primers extended with appropriate 5′ BamHI and 3′ SpeI restriction sites and subcloned into BamHI/NheI digested pGFP-N. Primer sequences are available upon request. The nuclear localization sequence (NLS)-GFP plasmid was produced by insertion of the 5′-GTACCATGCCCAAGAAAAAGAGGAAAGTTA and 5′-GATCTCCCAAAAAGAAAAGAAAGGTCG double-stranded oligonucleotides into the Asp718 and BamHI sites, respectively, at the 5′ and 3′ ends of GFP in pGFP-N. Otx2 fragments were then subcloned as described above. All constructs were sequenced on an automated ABI-prism sequencer (model 373, Applied Biosystems, Foster City, CA, USA). Beta-galactosidase expression plasmids (pCMV-beta and pCMV-beta-NLS) were kindly provided by E. Manet [16]. Otx2 sequences were inserted between the NheI and BglII sites of pCMV-beta-NLS, substituting to the SV40 NLS. Plasmid DNA was prepared on ethidium bromide–cesium chloride equilibrium gradients.

Cell culture

HeLa cells were cultured at 37°C, 5% CO2, and 95% humidity in Dulbecco Modified Eagle’s Medium (Invitrogen, Carlsbad, CA, USA) supplemented with 10% fetal calf serum (Perbio, Helsingborg, Sweden), 50 units/ml penicillin, 50 μg/ml streptomycin, and 2 mM l-glutamine (Invitrogen).

Reporter gene assay

The transcriptional activity of the Otx proteins was assayed using the region −66 to +68 of the IRBP promoter [17] cloned into the pSEAP2-basic vector (BD Biosciences, Palo Alto, CA, USA) as previously described [10]. The pSEAP2-Basic (promoter-less) and pSEAP2-Control (SV40 promoter) vectors were used as negative and positive controls, respectively. In standard assays, 105 HeLa cells per well were seeded in 24-well plates and transfected by the CaPO4 method with, unless indicated otherwise, 1 μg IRBP-SEAP, 0.25 μg GFP or GFP–Otx2 wild-type or truncated expression vector, and 0.1 μg pCMV-beta. After 40 h of incubation, secreted alkaline phosphatase (SEAP) and beta-galactosidase activities were measured. SEAP activity was normalized with beta-galactosidase activity. Normalized SEAP background in the absence of Otx2 expression vector was taken as one fold activation. Three independent experiments were performed in duplicate to generate each data.

Bacterial production of MBP–Otx2

The Otx2 coding sequence was subcloned into pMal-c2 plasmid (New England Biolabs, Beverly, MA, USA) and expressed in Escherichia coli BL21 cells. MBP–Otx2 fusion was affinity-purified onto an amylose resin according to the manufacturer’s instructions.

Nuclear HeLa extracts

Six hours before transfection, HeLa cells were seeded at 1.5×106 cells per 15 cm diameter dish, then transfected by the CaPO4 method with 10 μg of expression plasmid. Nuclear extracts were prepared 24 h after transfection according to Chelsky et al. [18] with some modifications. In brief, cells were harvested, washed twice with ice-cold PBS, and centrifuged at 500×g. The cell pellet was suspended in ten volumes of 10 mM Tris–HCl pH7, 10 mM NaCl, 3 mM MgCl2, 30 mM Sucrose, 0.5% (V/V) IGEPAL and a cocktail of anti-proteases (Complete, Roche Molecular Biochemicals, Indianapolis, IN, USA) and incubated for 10 min on ice. The lysate was centrifuged at 1,500×g for 10 min at 4°C and the nuclear pellet was suspended in two volumes of 20 mM Tris–HCl (pH 8), 400 mM NaCl, 1 mM EDTA, 0.1 mM EGTA, 1 mM dithiotreitol, 20% glycerol, and anti-proteases and incubated 15 min on ice. The nuclear lysate was cleared by centrifugation for 30 min at 10,000×g at 4°C then stored at −80°C. Protein concentration was determined using the protein assay kit (Bio-Rad Laboratories, Hercules, CA, USA).

Electrophoretic mobility shift assay

The oligonucleotides and their exact complement used were: wild-type IRBP IP1: 5′-CAGTAAGCCTTTAATCCTGTCT-3′; M1 mutant: 5′-CAGGCCTACTTTAATCCTGTCT-3′; M2 mutant: 5′-CAGTAAGCCTTGCCTACTGTCT-3′; and M3 mutant: 5′-CAGGCCTACTTGCCTACTGTCT-3′. Double-strand probes were 32P-radiolabeled with T4 polynucleotide kinase (New England Biolabs) and gel-purified. Ten fmol (2×104 cpm) radiolabeled double-stranded oligonucleotides were incubated for 30 min on ice with 5 μg of proteins, 1 μg of poly(dI–dC) as nonspecific competitor, in a buffer containing 25 mM Hepes pH 7, 50 mM KCl and 10% glycerol. Samples were resolved on 4% polyacrylamide non-denaturing gels. Electrophoresis was carried out at 4°C for 1 h at 15 V/cm in Tris–glycine buffer (40 mM Tris pH 8.8, 200 mM glycine). The gel was dried onto Whatman DE81 paper and autoradiographed.


Site selection was performed according to Pollock and Treisman [19] using a degenerated oligonucleotide 5′-CACGTAAGCTTCTGA(N)23AGTCGAATTCACTGC-3′ and complementary primers 5′-CACGTAAGCTTCTGA-3′ and 5′-GCAGTGAATTCGACT-3′ for PCR amplification. MBP–Otx2–DNA complexes were formed as in electrophoretic mobility shift assays (EMSA) using 0.5 ng probe with 0.5 μg protein and 0.1 μg poly(dI–dC), and separated on 6% polyacrylamide non-denaturing gels. The complexes were eluted in a buffer containing 0.1 M CH3COONH4, 50 mM Tris–HCl pH 8, 5 mM EDTA, and 0.5% sodium dodecylsulfate (SDS), and DNA was recovered by phenol extraction and precipitated with ethanol. PCR was carried out with 1 pg recovered template, 65 μM primers for 18 cycles (annealing at 52°C). PCR product was gel purified and quantified. Four rounds of selection were performed. The resulting selected PCR product was digested by EcoRI and HindIII and inserted into EcoRI/HindIII digested pBluescript SK+ (Stratagene, La Jolla, CA, USA) and 34 independent clones were sequenced.

Immunofluorescence and epifluorescence microscopy

HeLa cells were seeded on glass slides coated with 0.1% gelatin. Twenty-four hours after transfection, the cells were washed with PBS buffer and fixed with 10% formaldehyde. GFP constructs were observed by direct epifluorescence on a Zeiss Imaging system Axioplan 2 (Zeiss, Oberkochen, Germany) with a fluorescein isothiocyanate filter and analyzed with CoolSNAP software (Roper Scientific, Duluth, GA, USA).

For Western blot experiments, a rabbit polyclonal antiserum was raised against the synthetic peptide KSSPAREVSSESGT (Otx2 amino acids 113–126) coupled to Keyhole Limpet Hemocyanin by Covalab (Lyon, France) according to their standard protocol, and immunopurified on peptide sepharose column. This antibody was used at 1:500 dilution in PBS containing 0.1% Triton X100 and 5% milk powder.

Beta-galactosidase subcellular localization

Histochemical staining for beta-galactosidase activity in HeLa cells was carried out with 5-bromo-4-chloro-3-indolyl β-d-galactoside (X-gal). Fourty hours after transfection, the Hela cells were washed with PBS buffer and fixed for 10 min at 4°C in a 0.1 M-phosphate buffer (pH 7.3) containing 0.2% glutaraldehyde, 2 mM MgCl2, and 5 mM EGTA. The cells were washed three times in 0.1 M phosphate buffer at pH 7.3 containing 2 mM MgCl2, 0.01% (W/V) sodium deoxycholate and 0.02% IGEPAL. Staining was performed at 37°C in the above buffer supplemented with 1 mM spermidine, 5 mM K3Fe(CN)6, 5 mM K4Fe(CN)6, and 1 mg/ml X-gal. Staining was observed under an inverted Olympus microscope (Tokyo, Japan) equipped with a digital camera system and the TWAIN Viewfinder Lite software (Pixera, Los Gatos, CA, USA).


Transcriptional activity of Otx2 mutants and mapping of the transactivation domains

Construction of Otx2 mutants and transcriptional reporter assay

Mutations associated with human eye malformations comprise deletions, insertions, and transitions that result in C-terminal truncated and point mutants OTX2 proteins (Fig. 1a). To understand their effect on OTX2 activity and to simultaneously assign a biological function to the different conserved domains, we generated a series of deletions that covered the entire protein. Deletion borders were chosen using the motifs that are conserved among the three mammal Otx proteins, Otx1, Otx2, and Crx [3, 20, 21] (Fig. 1a). These include the Paired-class homeodomain [amino acid (aa) 37 to 97] flanked by two basic regions, b1 (aa 36–42) and b2 (aa 89–94), a glutamine stretch just C-terminal to the homeodomain (aa 95–101), a third basic region, b3 (aa 107–114), a WSP motif of unknown function (aa 150–159), and two repeated tails, OT1 (aa 255–267) and OT2 (aa 273–285) at the C terminus. Truncation mutants are described and numbered by their N- and C-terminal aa (wild-type protein is noted 1–289). In addition, the three reported point mutations resulting in a single amino acid change, R89G, P133T, and P134A, were created by site-directed mutagenesis. With human and mouse 289-aa major Otx2 isoforms sharing identical primary sequence, with the exception of a conservative threonine to serine change at position 115, all mutants were made using the mouse sequence that we previously cloned [10].
Fig. 1

Deletion and mutation mapping of Otx2 transactivation domains. a The structure of the Otx2 protein is drawn to scale. The hatched box represents the homeodomain. Black boxes b1, b2, and b3 are the stretches of basic amino acids, Q is a glutamine-rich region, and WSP box is a central conserved motif. OT1 and OT2 designate the tandem repeated conserved Otx tail motifs. Numbers indicate the aa position of truncations used in this study. The predicted proteins resulting from human spontaneous mutations are aligned underneath. FS1, 2, and 3 are frameshift mutations at positions 27, 39, and 155. respectively. Dotted lines show the translated sequence downstream the frameshifts. Q99X and Y179X are mutations creating a stop codon at indicated position. R89G, P133T and P134A are point mutations that change the identity of the indicated aa. These are represented by an open circle. b Transcriptional activity of full-length Otx2 protein and GFP-Otx fusions. HeLa cells were co-transfected along with a plasmid containing the −66 to +68 human IRBP promoter linked to a secreted alkaline phosphatase reporter cassette (pIRBP-SEAP), a plasmid expressing the indicated proteins and plasmid pCMV-beta expressing beta-galactosidase for normalization. Schematic representation of Otx proteins as above is on the left panel. All constructs are drawn so that their Otx2 part is vertically aligned. The GFP box represents the GFP peptide sequence (not drawn to scale) alone or fused either at the N- or C- terminus of Otx proteins as indicated. The right panel shows the corresponding transactivation activities. Normalized phosphatase activity is presented relative to pIRBP-SEAP. Error bars are standard deviation calculated from three independent experiments performed in duplicate. c The activity of Otx2 truncation and point mutant proteins was measured and represented as described above. Numbers in the left panel indicate the terminal amino acids of each tested mutant, besides its schematic representation. 1–106/247–289 indicates a peptide where region 1 to 106 is fused to the C-terminal region 247 to 289. The triple-point mutant in the homeodomain is numbered 1–289*, and the position of the mutations is symbolized by a star

The above constructs were assayed for transcriptional activation of the human IRBP proximal promoter, a known Otx2 responsive sequence [17]. Expression of mutant proteins was checked by Western blot (see Fig. 3) and found to be comparable. In HeLa cells, wild-type Otx2 activated this promoter strongly and specifically (Fig. 1b). N- or C-terminal GFP–Otx2 fusion proteins stimulated the IRBP promoter to the same extent than Otx2 protein, indicating that the GFP moiety did not perturb Otx2 activity. The same results were also obtained in human embryonic kidney 293T and mouse fibroblast 3T3 cell lines (not shown). Subsequent studies were, thus, carried out using N-terminal GFP fusions to facilitate transfection monitoring and detection of the assayed proteins.

C- and N-terminal regions account for maximal transcriptional activity

A panel of C- and N-terminal truncations was generated to map Otx2 transactivation domains (Fig. 1c). Stepwise deletion of the Otx tails OT2 (construct 1–267) and OT1 (1–247) decreased the activity to, respectively, 40 and 20% of the control. Further C-terminal trimming down to aa 106 had no additional effect, indicating that all the transactivation potential C-terminal to the homeodomain rely on the Otx tails. As this region is lost in all the truncated human mutant proteins (Fig. 1a), their transcription activity must be drastically reduced.

The deletion of the aa N-terminal to the homeodomain (construct 35–289) also lowered the activity to about half of the control, revealing transactivation potential of this region (Fig. 1c). It is noteworthy that the C-terminal truncations of this molecule (35–289 to 35–106) yielded the same activity profile as the previous series, albeit at a constant 50% efficiency. This strongly suggests that the N-terminal (1–35) and C-terminal (247–289) transactivation regions act independently. Construct 35–106, which only retains the homeodomain, was inactive showing that the DNA binding domain cannot activate transcription by itself. Although the 1–35 region is preserved in almost all human OTX2 mutants, it is likely that mutants FS1 and FS2, which lack a homeodomain and, thus, cannot bind to DNA, are totally inactive. By contrast, other mutants that have an intact homeodomain must retain this N-terminal-associated transactivation potential.

To test whether transactivation domains 1–35 and 247–289 together with the homeodomain could reconstitute Otx2 activity, we made an internal deletion between amino acids 106 and 247, which preserved only these regions. This construct, 1–106/247–289, exhibited twice the activity of its 1–106 counterpart, demonstrating intrinsic transactivation properties of the Otx tails (Fig. 1c). However, it failed to reconstitute full Otx2 activation, indicating that the missing 106–247 region, although devoid of activity, could cooperate with the Otx tails or help for optimal domain folding or spacing.

Transcriptional activity requires an intact homeodomain

As transactivation depends on Otx2 binding to its target site, it is expected to strictly rely on homeodomain integrity. Indeed, the N-terminal truncation of the first homeodomain alpha-helix (52–289), the second (67–289), or almost all of the homeodomain (90–289) totally abolished activity (Fig. 1c). Structural analysis [22] and functional studies [23] of other homeodomains have emphasized the critical role of three residues of the recognition helix in setting sequence specific protein–DNA contacts. These are valine 84, lysine 87, and asparagine 88 in the Otx2 protein. We changed these amino acids into phenylalanine, glutamic acid, and alanine respectively. The resulting triple-point mutant protein had no activity at all, confirming that DNA interaction is obligatory for Otx2 to stimulate transcription of the IRBP promoter. The R89G point mutant protein, although bearing intact N- and C-terminal transactivation domains, displayed only 25% of control activity. This residual activity indicated that although arginine 89 is important, its mutation does not completely abolish DNA interaction (see below). Nonetheless, the very reduced activity of R89G mutant observed here may be relevant to the phenotype of patients carrying this mutation. In contrast, mutants P133T and P134A exhibited full transcriptional activity (Fig. 1c). This result is consistent with their intact homeodomain and transactivation regions at both extremities of the protein.

In conclusion, Otx2 possesses two transactivation domains at its N- and C-termini whose activity relies on an intact homeodomain. It is interesting that either the lack of C-terminal domain or homeodomain mutations in human OTX2 mutants strongly reduce but do not totally abolish their transcriptional activity.

DNA binding properties of the wild-type and mutant Otx2 protein

We next asked whether DNA binding properties could be affected in human mutant OTX2 proteins. Previous studies have proposed that OTX2 could bind to the human tenascin-C promoter as a homodimer and that binding required the carboxy-terminal region [24]. Thus, DNA binding or dimer formation could also be impaired in the C-terminal truncated mutants Q99X, FS3, and Y179X that retain an intact homeodomain.

Sequence specificity

Although Otx2 sequence selectivity has not been thoroughly characterized, the protein can bind to the bicoid target site (bts) 5′-TAATCC-3′ of the drosophila hunchback promoter [25]. Otx2 DNA binding activity was probed in electrophoretic mobility shift assays using the IP1 motif of the human IRBP promoter [17], which contains a bts (Fig. 2a). Incubation with maltose binding protein MBP–Otx2 fusion protein yielded dose-dependent formation of a prominent high-mobility (H) complex and of a minor low-mobility (L) complex (Fig. 2b). The constant 30-fold ratio between both complexes suggested independent binding sites. A closer examination of IP1 sequence showed that besides the consensus bts (site 2, 5′-TAATCC) laid the related 5′-TAAGCC sequence (site 1, Fig. 2a). A possible involvement of this latter sequence was addressed by mutation analysis. Site 1 mutation (Fig. 2b, M1 probe) abolished formation of the L complex without affecting formation of the H complex. Mutation of the consensus bts (Fig. 2b, M2 probe) only left a weakly detectable H complex. No retardation was observed when both sites were simultaneously mutated (Fig. 2b, M3 probe). These results demonstrate that IP1 contains two Otx2 binding sites that have high (5′-TAATCC) and low (5′-TAAGCC) affinity. As complexes mobility does not change with increasing amounts of protein, we conclude that Otx2 binds each site independently as a monomer. The H complex is presumably made of an Otx2 monomer sitting on one site while the L complex has two Otx2 proteins sitting on each site.
Fig. 2

Otx2 can bind to different sequences. a The sequence of the IRBP IP1 probe used in EMSA experiments is shown. The canonical bts (5′-TAATCC) and its related motif (5′-TAAGCC) are in bold characters and underlined by a plain and a dotted line, respectively. b EMSA experiments using bacterially expressed Otx2 are presented. The schematized wild-type and mutant IP1 oligonucleotides (top) were labeled and incubated with 3 μg of maltose binding protein (MBP) alone (-) or with increasing amounts (0.1, 0.3, 1, and 3 μg, symbolized by triangles at the bottom) of MBP–Otx2 fusion. The complexes were separated by electrophoresis and revealed by autoradiography. IP1 site 1 and site 2 are symbolized by empty boxes and transversion mutants by crossed boxes. The position of low- (L) and high (H)-mobility complexes and free probe is indicated. c The sequence of sites selected by the SELEX method is presented with canonical TAATCC above and degenerated motifs below. A frequency diagram of base distribution around the TAATCC motif is presented at the bottom

The unexpected finding of Otx2 binding to TAAGCC sequence prompted us to better characterize the DNA sequences recognized by Otx2 using a SELEX approach. Starting from a pool of oligonucleotides randomized at a sequence of 23 nucleotides, we used EMSA to select for sequences with affinity for Otx2. We obtained the sequences of 31 independent aptamers after the fourth SELEX round (Fig. 2c). Although TAATCC sequence was prominently selected, 22% of the aptamers displayed different motifs, indicating flexible Otx2 sequence specificity. We found and extended preference at both ends of selected sites (Fig. 2c) and derived a consensus site: RCTAATCCCYY. To make it simpler, further analysis was carried out with M1 probe which contains the preferred TAATCC motif.

No sequence outside the homeodomain is required for DNA binding

Although integrity of the homeodomain is a requisite for DNA binding, one cannot exclude that other sequences cooperate with it for maximal DNA binding activity. To check this point, we performed EMSA (Fig. 3) with nuclear extracts from HeLa cells expressing various truncation mutants. Protein expression was monitored by Western blot analysis (Fig. 3, lower panel). C-terminal deletions preserving the homeodomain formed progressively shorter complexes of comparable intensity (Fig. 3, lanes 4–9). N-terminal deletion, up to aa 35, neither change the binding, indicating that Otx2 homeodomain is sufficient for full interaction with DNA (Fig. 3, lanes 10 and 11). Therefore, human mutant proteins Q99X, FS3, and Y179X (Fig. 1a) should bind normally to DNA. On the contrary, binding was abolished upon removal of the first homeodomain helix (construct 52–289, lane 12) and beyond (lanes 13 and 14). Thus, human OTX2 mutants FS1 and FS2 which lack the homeodomain are unable to bind DNA. Binding was also lost in the transcriptionally inactive triple-point mutant (lane 16), confirming the critical importance of the mutated residues. It is interesting that the R89G point mutant retained weak binding activity (Fig. 3, lane 17), in good correlation with its poor transcriptional activity. As glycine is a small uncharged residue, we conclude that arginine 89 side chain is involved in direct protein–DNA interaction. By contrast, mutations P133T and P134A that lie outside the homeodomain do not affect the DNA binding capacity (lanes 18 and 19).
Fig. 3

Determinants of Otx2 DNA binding activity. DNA binding of GFP fusions of the Otx2 truncation mutants collection and of Otx2 point mutants is shown. Upper panel: the indicated proteins, symbolized as in Fig. 1, were transiently expressed in HeLa cells. Middle panel: equal amounts of corresponding nuclear extracts were incubated with 32P-labeled IRBP IP1 M1 probe and then subjected to EMSA. The position of nonspecific complexes (ns), cleavage product (cp), and free probe is indicated. The dotted line indicates the position of the proline endopeptidase site in Otx2 protein. Bottom panel: Western blot analysis of the expressed proteins using an antibody raised against a peptide spanning Otx2 amino acids 114–128 is presented when possible. The molecular weight markers’ position is indicated

We noticed a recurrent complex (Fig. 3, lanes 4–8) of the same size as that formed with mutant 1–106 (lane 9). This was not observed with bacterially expressed MBP–Otx2 (see Fig. 2b) and it was less abundant in nuclear extract prepared from HEK293 cells (not shown). As it is detected in Western blot with an antibody against GFP and it binds to DNA, we interpreted it as a cleavage product that retains Otx2 homeodomain. A proline endopeptidase site present in Otx2 at position 110 could generate such a product as this enzyme is abundant in HeLa cells [26]. Its size, similar to mutant 1–106 as well as its truncation by the 35-aa N-terminal deletion (Fig. 3, lane 11) fully support such hypothesis. This cleavage product does not interfere with transcription assays (see Fig. 5b).

Our data together indicate that Otx2 DNA binding only requires an intact homeodomain. With this homeodomain having no intrinsic transcriptional activity, we conclude that, within the Otx2 protein, activation and binding domains are physically distinct and complement each other. Most of the human mutations affect either one or both domains resulting in partially or totally inactive proteins.

Otx2 nuclear localization relies on two independent determinants

Our in vitro analyses suggest that, with the exception of point mutants P133T and P134A, mutant OTX2 proteins are unable to activate transcription or to bind to DNA or both. However, another possible reason why they fail to exert their function could be their in vivo mislocation. Nuclear protein accumulation depends on both cytoplasm to nucleus transport and nuclear retention. Despite the clear nuclear localization of Otx2 [27, 28], the corresponding location signals have not been characterized. The putative basic motif inside and outside the homeodomain (Fig. 1a) could account for this property. To map Otx2 nuclear targeting signals, we used GFP fluorescence in cells expressing different GFP–Otx2 mutants. Whereas GFP alone was evenly distributed within the cell, the full-length Otx2 protein and all series of C terminus truncations appeared located to the nucleus (Fig. 4a). Human mutant proteins R89G, P133T, and P134A were also perfectly nuclear. This was also the case for the N-35 deleted mutants (Fig. 4a), indicating that the homeodomain (35–97) is sufficient for nuclear localization. However, neither N-terminal truncation of the first homeodomain helix (mutant 52–289) nor the second helix (67–289) did affect this subcellular distribution (Fig. 4b). The truncation of the first 89 aa that removed almost all the homeodomain and disrupted the b2 basic motif remarkably lead to an apparent nuclear targeting. This suggested that other determinants C-terminal to the homeodomain are also involved in accumulation of Otx2 in the nucleus and could substitute to missing or altered homeodomain. Supporting this idea, we could observe different consequences of the triple homeodomain mutation depending on the presence of the C-terminal region: while a fraction of the 1–106* protein pool appeared delocalized to the cytoplasm, full-length 1–289* protein remained nuclear (Fig. 4a). This indicates that aa C-terminal to 106 cooperate with the homeodomain for nuclear accumulation. Using a series of N-90 truncation mutants (Fig. 4b), we could restrain this nuclear distribution signal to the 117–146 portion of Otx2 protein. P133T and P134A mutations, which localize within it, surprisingly did not affect nuclear localization of the 90–146 fragment (Fig. 4c), indicating that the mutated proline residues are not critical for this activity.
Fig. 4

Two different Otx2 regions confer nuclear localization. a Sequences within the homeodomain contribute to tight nuclear GFP accumulation. The indicated series of truncation Otx2 mutants were expressed in HeLa cells and their subcellular localization appreciated by fluorescence microscopy. For each group of constructs, the localization was identical and only one representative mutant is shown. Protein aa numbers and symbols are as in Fig. 1. Full-length Otx2 protein is 1–289 and 35–97 is the complete homeodomain. The triple-point mutant is indicated by a star and the aa changes (V84F, K87E, and N88A). The shaded rectangle indicates the minimal common region in all mutants localized to the nucleus. b Identification of a determinant of nuclear accumulation outside the homeodomain. The subcellular localization of truncation mutants lacking partially or totally the homeodomain was determined as above. One representative mutant of each series with similar characteristics is shown. The minimal region required for nuclear accumulation is shaded and its primary aa sequence is presented at the bottom. Proline residues 133 and 134 are in bold and underlined. c Nuclear accumulation of the 90–146 region is not affected by mutations found in humans. The indicated polypeptides expressed in HeLa cells and their subcellular localization are presented. d Comparison of nuclear addressing potency of Otx2 sequences. HeLa cells were transiently transfected with plasmids containing the E. coli LacZ gene bearing the indicated in-frame insertions (x-box) under the control of the CMV early immediate promoter. The subcellular localization of beta-galactosidase was revealed by X-gal staining and bright-field microphotography. None, no insertion; NLS T SV40, the insertion corresponds to the strong basic prototypical nuclear localization sequence of the T antigen of Simian Virus 40 (PKKKRKV); Otx2 1–289, the full Otx2 coding sequence is inserted; Otx2 117–146, the sequence inserted encodes the 30-aa motif identified above, encompassing Otx2 117–146 region

Then, we set out to characterize this new putative nuclear accumulation signal. Contrary to classical NLS [29], the 117–146 domain lacks basic amino acids but has a remarkably high proportion (46%) of serine and threonine residues (Fig. 4b). Because GFP (alone or in fusion with small peptides) is known to freely diffuse across nuclear pores, we could not exclude that nuclear fluorescence observed with the 117–146 GFP construct resulted from nuclear retention and accumulation of this diffusible protein. To address this point, we used a 117–146 fusion with E. coli beta-galactosidase which requires active import to enter the nucleus [16]. Contrary to SV40 T antigen NLS used as a control [29], Otx2 117–146 domain did not confer nuclear localization to the beta-galactosidase, whereas full-length Otx2 did (Fig. 4d). This demonstrated that the sole genuine Otx2 NLS resided in the homeodomain and that the 117–146 region rather acted as a nuclear retention domain, which could cooperate with the homeodomain NLS to stabilize Otx2 nuclear localization.

Search for dominant negative effects

We have shown that the human OTX2 mutant proteins Q99X, FS3, and Y179X, although devoid of transcriptional activity, have an intact homeodomain and, thus, can efficiently bind to DNA. Patients carrying these mutations are heterozygous. This raises the possibility that, in vivo, the mutated protein acts as a dominant negative towards its normal counterpart, through competitive binding of target sites or interacting partners. To test for any possible interfering activity of C-terminal truncated proteins, the N terminus fragment containing the homeodomain (1–106) was expressed together with Otx2 (Fig. 5a). No inhibition of the transcriptional activity of Otx2 was observed, although this fragment was shown to be nuclear (Fig. 4a) and to bind to DNA (Fig. 3). A dose–response experiment showed that even a molar excess of this portion did not influence the activity of the wild-type protein (Fig. 5b). Thus, one cannot put forward a dominant negative effect of the shortest truncated OTX2 mutants FS1, FS2, and Q99X.
Fig. 5

The Otx2 region containing amino acids 117–146 exert dominant negative effects. a Activation of the IRBP promoter by full-length GFP–Otx2 was challenged by equimolar amount of the indicated constructs (left panel, symbols as in Fig. 1). HeLa cells were cotransfected with 0.5 μg GFP–Otx2 expression plasmid and equimolar amounts of plasmid encoding the indicated truncation mutants. The corresponding activity of the SEAP reporter gene was represented as the percentage of residual activation on the right panel. Error bars are standard deviations. Extra NLS from SV40 T antigen that have been added to force nuclear localization of otherwise cytoplasmic Otx2 fragments are symbolized by the white N letters in black circles. b Specificity and dose–response activity of a dominant negative mutant. HeLa cells were transfected with the IRBP-SEAP reporter along with 0.5 μg GFP–Otx2 expression plasmid (left) or SV40-SEAP vector (right) and the indicated amounts (0–3 μg) of plasmids expressing the Otx2 fragments 1–106 or 90–146 in fusion with GFP. SEAP activity was expressed as percent of control. c Modification of the subcellular localization of Otx2 fragments. The indicated regions of Otx2 protein were transiently expressed in HeLa cells either as fusions with GFP (left) or with a modified version of GFP flanked by two NLS from SV40 T antigen (right). Their distribution was observed by fluorescence microscopy

Modulation of Otx2 transcriptional activity by the 117–146 region

We next tested the C-terminal Otx tails region. We reasoned that as it accounts for most of the transcriptional activity, it might titrate out cellular co-activators or general transcription factors. It was surprising that no inhibitory effect was observed (Fig. 5a). In contrast, the 90–289 construct and C-terminal deletions down to minimal 90–146 region reduced the activity of the wild-type protein to about 50% of controls (Fig. 5a). This effect was dose dependent and specific to the Otx2 responsive IRBP promoter (Fig. 5b). The minimal 90–146 region carrying mutations P133T or P134A did exert a comparable dominant negative effect (Fig. 5a), once again suggesting that proline 133 and proline 134 are not critical residues. Further deletion of the amino acids 117–146 abolished this effect (Fig. 4a), which could be explained by the retention of these latter mutants, as well as the Otx tails fragment, in the cytoplasm (Figs. 3a and 5c). We forced their nuclear localization by fusing them to a GFP variant carrying two strong nuclear localization signals. Despite an efficient nuclear import (Fig. 5c), none of them displayed any further inhibitory activity (Fig. 5a). Thus, the only region of the Otx2 protein that autonomously exhibited a dominant negative effect on its own transcriptional activity was peptide 117–146. Located between the conserved basic b3 and WSP motifs, it contains the nuclear retention signal and a series of potential phosphorylation targets. The proline residues 133 and 134 that are mutated in human patients do not appear essential for its activity.


To characterize the molecular defects of human OTX2 mutants, we have mapped the functional domains of Otx2 transcription factor using extensive deletion analyses of GFP-tagged mutants. Our data show that Otx2 has a typical transcription factor modular architecture with discrete and separable DNA binding and transactivation domains, the latter having a split tripartite structure. Both corresponding activities are either separately or together affected in human mutant OTX2 proteins, leading to decreased or absent activity but not to dominant negative effect. This study also unravels an unexpected nuclear retention domain, the activity of which can interfere with the transcriptional function of the protein.

DNA binding and nuclear localization

An integral Otx2 homeodomain is necessary and sufficient for DNA binding activity. Deletion of its N-terminal part totally abolishes DNA binding like the truncation of its C-terminal part in Crx does [30]. R89G mutation, which causes bilateral microphtalmia in humans [14], only decreases binding and transactivation to about 20% of control without affecting nuclear localization. This residue, which makes hydrogen bonds with the first T of the TAATCC motif in the related Pitx2 homeodomain–DNA complex [31], appears also crucial to stabilize Otx2–DNA interaction. Contrary to carriers of mutations that lead to C-terminal truncation of OTX2 transactivation domain causing total anophtalmia, carriers of this R89G mutation still develop reduced eyes. Residual transcriptional activity in mutant R89G protein could account for this difference of phenotype.

The DNA binding specificity of Otx2 remains to be firmly established. Many putative target genes have been revealed by classical [8, 32] and genome-wide analyses [33], but very few Otx2 in vivo DNA target sequences have been experimentally validated [34]. Bicoid-type homeodomains preferentially bind to the 5′-TAATCC motif [23], which was confirmed here, with extended 5′ and 3′ specificity, making a consensus 5′-RCTAATCCCYY-3′ binding site. Our selection of more distantly related motifs also raises the possibility that Otx2 could recognize a wider spectrum of lower-affinity sequences in vivo. This is in perfect agreement both with our observation of a weaker binding to the 5′-TAAGCC sequence in IRBP promoter (Fig. 2b), and with structural analysis of the Pitx2 homeodomain–DNA complex which suggest that K50 homeodomains could sample multiple DNA binding sites [31].

The physiological protein–DNA stoechiometry is also an open question. The binding properties of the Paired-class proteins have been previously investigated [35]. While Paired homeodomain showed strong cooperative dimerization, the Otd class, represented by an altered-specificity Paired homeodomain (K50-Prd-HD) and by the Gsc homeodomain, exhibited the weakest cooperativity. These data were obtained with naked homeodomains which might behave differently from full-length proteins. Moreover, a search for optimal Bcd and Otd homeodomain binding sequences [36] yielded non-palindromic site after five rounds of selection, suggesting monomeric binding. Our data, demonstrating monomeric binding of Otx2 to DNA, provided a single site probe is used, strengthen this conclusion. Others have shown that Otx2 could dimerize, even in the absence of DNA, through homeodomain or C-terminal interactions [24, 37]. These interactions were observed using high concentrations of bacterially and in vitro expressed Otx2 protein, although the protein was found to be monomeric in solution during purification. In addition, the presence, besides the canonical TAATCC sequence, of a TAACCC motif, on the complementary strand of tenascin-C promoter [24] might explain the binding of two Otx2 polypeptides per DNA molecule. In this study, we show equivalent binding levels with full-length protein and mutants truncated at their N- and C-termini (Fig. 3). This precludes that Otx2 dimer formation is a prerequisite for DNA binding and, hence, excludes the outcome of inactive heterodimers as a result of human Otx2 gene mutations.

The Otx2 homeodomain has also intrinsic NLS activity. It is remarkable that this is not modified by human R89G mutation. Even triple-point mutations that abolish DNA binding only partially affect nuclear import, indicating that both activities rely on different residues or structural features of the homeodomain. Our data clearly show that the b2 motif of the Otx2 homeodomain has no intrinsic NLS activity (Fig. 4b). This b2 motif might be an element of a multipartite NLS [29]. This situation differs from that reported for the Crx gene where b2 mutations cause nuclear leakage [38].

Independent transactivation domains

Our results delineate two transactivation domains at both ends of the Otx2 protein. The duplicated Otx tails are essential transactivation domains, in good agreement with what has been reported for Crx and Otx1 [21, 30, 39]. Human mutations that cause only Otx tails truncation are functionally equivalent to those that block protein expression, indicating that they must mediate essential Otx2 biological activity. The truncation of each Otx tail causes a similar decrease of activity, indicating that each of them can act independently as an elementary transactivation unit. Their lack of dominant negative activity, even when targeted to the nucleus, suggests that their molecular targets are not rate limiting, as it is the case for general transcription factors [40].

The present work also uncovers a novel Otx2 transcriptional domain located in the first 35 N-terminal amino acids. Although the sequence of this region is conserved, this function was neither described in Otx1 [39] nor in Crx [30, 38]. The presence of several independent transactivation domains in Otx2 suggests that the protein can integrate signals from simultaneous activating pathways.

Serine–threonine-rich region with mixed activities

This study revealed a new unexpected nuclear retention function to the serine-rich region 117–146. It is interesting that this region has also a dominant inhibitory effect for Otx2 transactivation. Human spontaneous proline mutations in this domain appear to be neutral towards both activities suggesting that it is not under strong structural constraint. The biased amino acid composition of this motif (46% serine, threonine) suggests that it might undergo specific post-translational modifications, such as nuclear phosphorylation. Nuclear retention and stimulation of activity upon phosphorylation of a serine-rich domain has been reported for thyroid hormone receptor alpha-1 [41]. A serine-rich motif was also shown to contribute to the minimal transactivation domain of the Crx partner NRL [42]. The dominant negative activity of the 117–146 peptide could be explained by competitive saturation of nuclear platforms where Otx2 has to be recruited. This region could alternatively compete with the wild-type protein for post-translational modifications required for full transcriptional activity. Similar modifications could be specifically shared by other homeodomain proteins as the Otx2 117–146 fragment also behaves as a dominant negative towards Pax6 protein activity (S. Saule, personal communication) but does not interfere with SP1-dependent activation of the SV40 promoter (Fig. 5b). Further studies are required for a better biochemical characterization of this domain.

Interpretation of human mutations

Among the eight spontaneous mutations in human patients with severe eye malformations, only frameshift mutation FS1 creates a stop codon in the second Otx2 coding exon that could lead to mRNA nonsense-mediated decay [14]. The resulting haploinsufficient situation would not be different from that expected if FS1 protein is expressed. Mutation FS1 together with FS2 indeed yield truncated proteins without homeodomain that cannot bind DNA nor activate transcription. These mutants must be totally inactive. Q99X stop mutation, frameshift mutation at position 155, and Y179X stop mutation are expected to resemble our 1–106, 1–161, and 1–247 deletion mutants, respectively. All show normal DNA binding and weak transactivation, and none of them can exert dominant negative effect. These mutants may, therefore, behave as hypomorphs. The homeodomain R89G point mutant has normal transactivation but weak binding potential, resulting in a poorly active protein in vivo. Finally, two point mutations, P133T and P134A, fall into the nuclear retention and dominant negative domain, neither affecting those activities nor DNA binding, transactivation and nuclear localization. This raises the question whether these mutations are causal to the phenotypes of carriers [14]. P133T was indeed present in two unaffected relatives of the patient where it was found, and the mother of patient carrying P134A mutation had similar ocular malformation but lacked this mutation. Although we cannot rule out perturbed protein–protein interactions affecting the region of aa 133–134 of the OTX2 protein, it is likely that these mutations are neutral with respect to Otx2 function and are not responsible for the eye phenotype of their carriers.

Implications for genetic counseling

Having been causally linked to eye malformations, Otx2 mutations will now be searched in families affected by ocular defects. Our analysis identifies essential functional domains of the Otx2 protein and, hence, allows one to foresee the consequences of human mutations. Most of truncations that lead to the loss of both Otx tails are likely to compromise development, while point mutations may have various effect depending on their position. Such mutations can be reproduced and readily assayed using the biological and biochemical tests presented here.


We thank Jérôme Lacroix who contributed to this work as an undergraduate student, Evelyne Manet and Alain Sergeant for helpful discussions, and Charlie Scutt for improving the manuscript. NF is a recipient of a fellowship of the French Ministry of Research and Education. This work was supported by grants from the CNRS, the Retina France association, and the Comité du Rhône of the Ligue Nationale contre le Cancer.

Copyright information

© Springer-Verlag 2006