Regulation of Expression within a Gene Family THE CASE OF THE RAT γB- AND γD-CRYSTALLIN PROMOTERS

The six closely related and clustered rat γ-crystallin genes, the γA- to γF-crystallin genes, are simultaneously activated in the embryonic lens but differentially shut down during postnatal development with the γB-crystallin gene, the last one to be active. We show here that developmental silencing of the γD-crystallin promoter correlates with delayed demethylation during lens fiber cell differentiation. Methylation silencing of the γD-crystallin promoter is a general effect and does not require the methylation of a specific CpG, nor does methylation interfere with factor binding to the proximal activator. In later development, the γD-crystallin promoter is also shut down earlier by a repressor that footprints to the −91/−78 region. A factor with identical properties is present in brain. Hence, a ubiquitous factor has been recruited as a developmental regulator by the lens. All γ-crystallin promoters tested contain upstream silencers, but at least the γB-crystallin silencer is distinct from the γD-crystallin silencer. The γ-crystallin promoters were found to share a proximal activator (the γ-box; around −50), which behaves as a MARE. The γB-box is recognized with much lower avidity than the γD-box. By swapping elements between the γB- and the γD-crystallin promoter, we show that activation by the γB-box requires a directly adjacent −46/−38 AP-1 consensus site. These experiments also uncovered another positive element in the γD-crystallin promoter, around −10. In the context of the γD-crystallin promoter, this element is redundant; in the context of the γB-crystallin promoter, it can replace the −46/−38 element.

The mammalian genome contains a large number of gene families, which encode related proteins with similar structure and function, yet are optimized for a particular developmental and differentiation stage. The pattern of expression of gene family members varies between families. For example, in the ␤-globin gene family, the paradigm of a clustered gene family, expression switches between members such that only one or two genes are active at the same time (for review, see Ref. 1). In contrast, the six clustered and closely related members of the ␥-crystallin gene family (the ␥Ato ␥F-crystallin genes), which encode abundant structural proteins of the vertebrate lens, are all simultaneously active in the embryonic lens but switched off individually during postnatal development (2)(3)(4)(5). In the rat, at 3 months of age the ␥B-crystallin mRNA is still present at 90% of the level at birth, whereas the transcript level from the ␥D-crystallin gene has dropped to 60%, and those of the ␥Eand ␥F-crystallin genes have dropped to 5% of the level at birth (4). As lens cells do not die and as the younger lens cells overlay the older cells, the consequence of this pattern of gene expression is the creation of a ␥-crystallin gradient across the eye lens, which correlates inversely with the water gradient. This gradient in turn sets the gradient of refraction across the lens and thereby prevents optical aberration.
The mechanism of the developmental regulation of the ␥-crystallin gene expression is not known. There is a strong negative correlation between the methylation state of a ␥-crystallin gene promoter region and gene activity, suggesting that DNA methylation, or rather lack of DNA demethylation, is involved in silencing the genes (6). Differential expression or availability of transactivating factors is also likely to be causally involved in developmental regulation of expression. It is generally assumed that the closely related ␥A-F-crystallin genes share a common regulatory element that specifies the lens specificity of these genes. The prime candidate for such an element is the palindromic sequence (here denoted the ␥-box) located upstream of the TATA box (Fig. 1). Mutations of this sequence abolish promoter activity in transfection studies (7,11,12). Furthermore, Goring et al. (13) have shown that a pentamer of the mouse ␥F-crystallin ␥-box sequence directs lens-specific expression in transgenic mice. The expression of this construct, however, was restricted to the embryonic lens nucleus, and it was suggested that the wider range of developmental expression of the mouse ␥F-crystallin promoter is determined by upstream enhancers (13,14).
The study of the regulatory mechanisms of crystallin gene expression is complicated by the peculiar mode of growth of the lens; the lens epithelial cells differentiate to lens fiber cells at the equator of the lens. The fiber cells of a late developmental stage but at an early differentiation state thus overlay fiber cells of an earlier developmental stage but at a later differentiation state. The lens is thus a mixture of cells at different developmental and differentiation stages. To obtain a fiber cell at a specific developmental and differentiation stage, we have made use of an in vitro differentiation system. In this system, the monolayer of epithelial lens cells, still attached to the lens capsule, is cultured in the presence of bFGF, 1 which induces the differentiation of lens epithelial cells to lens fiber cells (Ref. 15; for review, see Ref. 16). The lens fiber cells follow the course of differentiation also seen in vivo, including the typical changes in morphology and the accumulation of the various crystallins. The lens epithelial cells are aware of their developmental age and differentiate to fiber cells corresponding to that developmental age (17,18). When explants are taken from newborn rats, copious accumulation of ␥-crystallin is seen after about 10 days of in vitro culture (19 -21). Lens explants isolated from older rats differentiate more slowly in vitro than those from younger rats and accumulate less ␥-crystallin mRNA and protein (18,21); in differentiating explants from 10-day-old rats, the ␥-crystallin mRNA levels are only 1% of that seen in newborn explants, 2 and below the level of detection in explants from 14-day-old rats (21).
In a previous study (7), we analyzed the course of activation of the ␥D-crystallin promoter during the in vitro differentiation of rat lens explants isolated from newborn rats. Demethylation of this promoter occurs within the first 2 days of in vitro differentiation, long before activation of the endogenous gene. The pulse of activity of the endogenous gene, between days 10 and 12 (21), was suggested to be regulated by the balance of activity of a transactivating factor binding the ␥-box, first detected around day 6, and of a silencing factor, which appears around day 10 (7). To investigate developmental changes in these regulatory interactions, we have now followed the activation of the ␥D-crystallin promoter during in vitro differentiation of lens epithelial explants isolated from 10-day-old rats. We show here that in explants from these older rats, promoter demethylation is delayed, whereas the silencing factor appears earlier. We have further compared the ␥D-crystallin promoter with the ␥B-crystallin promoter, the promoter with the most extended developmental expression. We show that the ␥-boxes of the ␥Dand ␥B-crystallin promoters, which resemble a Maf recognition element (MARE; Refs. 22 and 23), are recognized by the same factor, possibly a Maf protein, but that the affinity of the ␥B-box for this factor is much lower than that of the ␥D-box. Activity of the ␥B-crystallin promoter requires interaction with an AP-1 binding site directly downstream of the ␥B-box. Like the ␥D-crystallin promoter, the expression of the ␥B-crystallin promoter is subject to silencing, but the silencing factor differs from the one that represses the ␥D-crystallin gene.

EXPERIMENTAL PROCEDURES
Cell Culture-Lens epithelial explants from newborn or 10-day-old (as indicated) Wistar rats were obtained essentially as described (24). Rat lenses were isolated in Medium 199 (Life Technologies, Inc.). The lens capsule together with the anterior monolayer of epithelial cells were peeled off the fiber cell mass and pinned down on a 3.5-cm Petri dish. Explants (three per dish) were cultured as described previously (7). Basic FGF (a kind gift from Scios, Inc., Mountain View, CA) was added to a final concentration of 25 ng/ml, and the cells were cultured for the indicated period prior to transfection.
Isolation of Chromosomal DNA and Ligation-mediated Polymerase Chain Reaction (PCR)-Isolation of chromosomal DNA from lens explants and ligation-mediated PCR was performed as described previously by Dirks et al. (7).
In Vitro Methylation of DNA-CpG-methylation of DNA using SssI methylase was essentially according to the manufacturer's protocol (New England Biolabs). Total reaction time was 4 h, whereby addition of enzyme and S-adenosylmethionine to the reaction mix was repeated after 2 h. Completeness of methylation was tested by a pilot digestion using ThaI and electrophoresis through an agarose gel.
DNA Transfection, Chloramphenicol Acetyltransferase (CAT) Assay, and ␤-Galactosidase Assay-Plasmid DNA was isolated according to the alkali lysis procedure (27) in conjunction with either the Wizard Maxiprep System (Promega) or CsCl gradient centrifugation (27). DNA was transfected to the lens cells using either Lipofectamine Reagent (Life Technologies, Inc.) or the PDS-1000/He Biolistic Particle Delivery System (Bio-Rad). When cells were transfected using Lipofectamine, per dish 2.0 g of CAT reporter construct and 0.25 g of CMV/␤-gal construct (29) were transfected to the cells according to the manufacturer's protocol. Using the Biolistic Particle Delivery System, 0.5 g of CAT reporter construct and 0.125 g of CMV/␤-gal construct was coated on 1-m gold particles and bombarded on the cells at 450 psi helium. After culturing for 3 more days in the presence of 25 ng/ml bFGF, the cells were harvested in 100 l of reporter lysis buffer (25 mM bicine, pH 7.8, 0.05% Tween 20, 0.05% Tween 80) per dish, and vigorously shaken for 10 min. The cell debris was pelleted in an Eppendorf centrifuge. To determine transfection efficiency, 20 l of the supernatant was used to assay for ␤-galactosidase activity (29). The remainder of the supernatant was heated for 15 min at 65°C to inactivate cellular deacetylases. From the supernatant, 20 l was used to assay for CAT activity as described by Gorman et al. (30) or using the Quan-T-CAT system (Amersham). Transfections were done in duplo or triplo, and two DNA isolates from each construct were tested in independent experiments.
Electrophoresis Mobility Shift Assay (EMSA)-Nuclear extracts were prepared as described previously (12). EMSAs were performed essentially as described (31,32). DNA restriction fragments were size-fractionated through a native 6% polyacrylamide gel and isolated by electro-elution using a Bio-Trap apparatus (Schleiger and Schuell), according to the manufacturer's protocol. Approximately 0.1-0.5 ng of end-labeled probe (10,000 -20,000 cpm) was added to 5-10 g of nuclear extract (5 l, final concentration of 100 mM NaCl) and either 1.0 g of poly(dGdC⅐dGdC) or 1.0 -3.0 g of poly(dIdC⅐dIdC), as indicated, in binding buffer (final concentrations 20 mM HEPES, pH 7.9, 10 -50 mM KCl as indicated, 1 mM EDTA, 1 mM DTT, 4% (v/v) Ficoll) in a total volume of 20 l. The reaction mixture was left for 10 min at room temperature, loaded on a pre-run 4% (w/v) polyacrylamide gel in 0.25 ϫ TBE (1 ϫ TBE ϭ 89 mM Tris-HCl, 89 mM boric acid, 2.5 mM EDTA), which then was run for 2 h at 10 volts/cm. The gel was dried and exposed to a Fuji AX film overnight with one intensifying screen.
In Vitro Footprint Analysis-The appropriate DNA fragment was 32 P-labeled at one end. An aliquot (600,000 cpm) was methylated using dimethylsulfate (DMS) essentially as described (33,34). The methylated probe was incubated with 150 -200 l of nuclear extract (150 -400 g of protein) for a preparative gel retardation assay (analytical assay 30 -40-fold scaled up). The complexed and the free probe were visualized by autoradiography overnight. The DNA was cut out of the gel, isolated by electro-elution as described above, cleaved by piperidine (final concentration 10% (v/v)), and size-fractionated in a 15% sequencing gel. The gel was dried, and exposed to a Fuji AX film for 18 -72 h.

Demethylation of the ␥D-Crystallin
Promoter Region-In explants from newborn rats, the ␥D promoter region is fully demethylated between day 1 and 2 of in vitro differentiation (Ref. 7; see also Fig. 2A). To test whether promoter demethylation still occurs in 10-day-old rat explants, in which the level of expression of the ␥D promoter is about 50-fold lower, the state of methylation of the genomic ThaI site at position Ϫ13 of the ␥D promoter was followed during in vitro differentiation. In a parallel experiment, the methylation state of this ThaI site in explants from newborn rats was tested. In newborn rat explants, virtually complete demethylation of the ThaI site was found after 2 days of culture, in agreement with the results of Dirks et al. (7). In contrast, demethylation of this site was significantly slower during differentiation of explants from 10day-old rats ( Fig. 2A). Even after 5 days of differentiation, demethylation was only 65% complete.
To determine the effect of DNA methylation on ␥D promoter activity, a ␥D(Ϫ73/ϩ10)CAT fusion gene was methylated using CpG methylase and transfected into explanted lens cells. We found a strongly reduced activity of the methylated construct; activity was only 1% of that of the unmethylated promoter and within background levels (Fig. 2B). Although the CAT coding region is also methylated in these experiments, other studies have shown that DNA methylation does not impede elongation (35)(36)(37) and that methylation of the CAT coding sequence does not affect transient expression (6,38). Hence, the effect of CpG methylation is likely to block ␥D promoter activity, although from our own experiments, we cannot rule out an aspecific effect.
The ␥D promoter contains a CpG site in its proximal activator, the ␥-box, located around Ϫ50 (see also Fig. 1). To test whether methylation of the ␥-box element is sufficient to block binding of the cognate activating factor in vitro, the binding of rat lens nuclear extract factors to a methylated promoter fragment was compared with that to a nonmethylated fragment in an EMSA. Complex formation with the methylated fragment was reduced when compared with that of the unmethylated fragment, but not abolished (complex D1; Fig. 2C). This was confirmed by the fact that the methylated fragment competed for the activator complex as efficiently as the unmethylated fragment itself. In the EMSAs using the methylated ␥D promoter fragment, an additional band is seen (complex D2; note that this complex migrates slower than the faint aspecific com-plex seen in some of the other lanes). This band could represent binding to the methylated DNA by general M CpG binding proteins.
These data suggest that methylation of the ␥-box is not sufficient to suppress promoter activity. We therefore tested the effect of methylation on the activity of mutant constructs, lacking either the CpG site at Ϫ50 or the CpG sites between Ϫ20 and Ϫ10. The activity of these mutant ␥D promoters, when methylated, was also in the background range (Fig. 2B). Similarly, the activity of the ␥C promoter was also very low when methylated. Together, our results indicate that the reduction in ␥-crystallin promoter activity by methylation is a general effect and not due to methylation of a specific site.
Appearance of Trans-acting Factors during in Vitro Lens Cell Differentiation-We have previously proposed that the differentiation stage-specific expression of the ␥D-crystallin gene during fiber cell differentiation was regulated by the phased appearance of first an activating and then a silencing factor (7). The reduced activity of the ␥D promoter in explants of 10-dayold rats could be due to a changed expression profile of these factors. Therefore, the activities of the ␥D(Ϫ73/ϩ45)CAT fusion gene and of a silencer-tkCAT construct were followed during the course of differentiation of 10-day-old explants. The ␥D(Ϫ73/ϩ45)CAT construct was active at all stages of differentiation, with a maximum around day 12 (Fig. 3A). The timing of up-regulation of the ␥D activating factor in these explants is very similar to that in explants from newborn rats (see Ref. 7). However, the activity of the ␥D(Ϫ73/ϩ45)CAT construct in 10-day-old explants was around 50% of that in explants from newborn animals (data not shown), indicating that the level of the activating factor is decreased in the older explants.
Rather different results were obtained when the presence of the silencing factor was assayed for; the construct containing four copies of the silencing element in front of the HSV tk promoter was inactive even in early differentiated cells from 10-day-old rats (Fig. 3B), indicating that the silencing factor is present throughout differentiation of these cells. In contrast, in explants from newborn rats, silencing activity was maximal only after 10 days of in vitro differentiation (Ref. 7; see also Fig. 5B). The level of silencing in fiber cells from 10-day-old rats was not significantly different from that during late differentiation of cells from newborn animals. The earlier appearance of the silencing factor might well explain the reduced activity of the ␥D gene in the 10-day-old explants.
The ␥D-Crystallin Silencing Factor-The ␥D silencer was originally found by chance, when testing the effect of the Ϫ84/ Ϫ71 G/C-rich region conserved among the ␥-crystallin promot-FIG. 1. Sequence alignment of the 5-flanking regions of the six rat ␥-crystallin genes. The sequence of the ␥-box region of the mouse ␥F-crystallin promoter is shown for comparison. Only the ␥D sequence is shown in full; for the other sequences, only differences are specified. Dashes indicate gaps introduced to optimize alignment. The TATA-box is shown in bold. Transcription start sites are indicated by arrows. Asterisks mark residues involved in factor binding in lens cells, as demonstrated by in vivo footprint analysis (7). The activator element at position Ϫ57 to Ϫ46 of ␥D-crystallin promoter, the ␥-box, is indicated as suggested by Peek et al. (8) as is the silencer region. Note that the nucleotide sequence of the mouse ␥F-crystallin ␥-box is identical to that of the rat ␥D-crystallin ␥-box except for the A, which is present in the rat ␥Aand ␥C-crystallin ␥-box sequences as well. Also note that the ␥B-crystallin equivalent of the ␥-box, and the region directly downstream, contains the most nucleotide changes relative to that of the ␥D-crystallin, and that the ␥Eand ␥F-crystallin promoters lack the G/C-rich Ϫ10 region present in the ␥D-crystallin promoter. Sequences and their alignment are according to Den Dunnen et al. (9); the mouse ␥F-crystallin promoter sequence is from Lok et al. (10).
ers. This region suppresses gene activity in non-lens cells such as retina, skin, and brain (12). In vivo footprinting in lens cells, however, showed that nuclear factor contacts extended further upstream to Ϫ88 (7). In view of its importance in the developmental shut down of the ␥D gene, we have reexamined this region. Methylation interference footprinting showed a contacted region between Ϫ91 and Ϫ79 in the upper strand and between Ϫ90 and Ϫ78 in the lower strand in both a lens and a brain nuclear complex, suggesting the presence of the same factor in lens and brain cells (Fig. 4, A and B). This was further confirmed by measurement of the molecular weight of the factors in an EMSA-based method (39), in which the mobility of the complexes in gels of different polyacrylamide concentrations were compared. The molecular masses of both the lens and brain complexes, including the oligonucleotide probe, were estimated at 105 kDa (Fig. 4C). These results strongly suggest that the lens and brain factors are the same and bind the ␥D silencer element at position Ϫ91 to Ϫ78 in both tissues. Apparently, this ubiquitous, or at least not lens-restricted, factor has been recruited by the lens to function in the differentiation and developmental control of the ␥D-crystallin promoter.
A synthetic copy of the Ϫ91/Ϫ78 element silenced the heterologous tk promoter by about 65% (Fig. 5A; ␥DS1), which corresponded to the silencing activity of a larger promoter fragment (␥D), whereas the conserved G/C-rich region (Ϫ84/ Ϫ68; ␥) silenced by about 25%. Thus, the Ϫ91/Ϫ78 element is the silencer element within the ␥D promoter. The in vivo aclenses. Binding was in the presence of 50 mM KCl and 50 ng/l poly(dGdC⅐dGdC). Complex D1 represents the ␥D-crystallin activator complex (␥-box complex), as confirmed by methylation interference footprint analysis using the Ϫ73/ϩ45 fragment (results not shown). Complex D2 might represent binding of M CpG-binding protein, as it is found only with the methylated probe. Specific competitor DNA was added in a 100-fold molar excess (right two lanes) .   FIG. 2. Demethylation of the ␥D-crystallin promoter during development. A, analysis of the methylation state of the genomic ThaI site at position Ϫ13. Lens explants from both newborn and 10-day-old rats were cultured in the presence of bFGF and harvested at several stages during differentiation. Chromosomal DNA was isolated as described previously (7) and digested with both ThaI (cuts at Ϫ13) and Sau3A (cuts at Ϫ23). The DNA was amplified by ligation-mediated PCR using three primers in succession (from ϩ85/ϩ67, ϩ42/ϩ21, and ϩ42/ ϩ19; for details see Ref. 7) and visualized by electrophoresis and autoradiography. PCR products from methylated and unmethylated DNA are indicated. Both the autoradiograph and the quantitated data are shown. B, activity of methylated (mutant) ␥Dor ␥C-crystallin promoters. Wild-type and mutant Ϫ73/ϩ10 ␥D-crystallin promoter constructs or the Ϫ70/ϩ28 ␥C promoter construct were in vitro methylated using CpG methylase (SssI methylase). Methylated and mock-methylated constructs were transfected to explants pre-cultured for 10 days in the presence of bFGF. The explants were cultured for three more days before harvesting (see "Experimental Procedures" for details). Activities of the non-methylated ␥D and ␥C promoter constructs were set at 100%; the activity of ␥DB2 or ␥DF is given relative to that of ␥D. The bars indicate the standard deviation. C, factor binding to the CpG-methylated ␥-box. Methylated and unmethylated ␥D (Ϫ73/ϩ45) fragments were used as probes, as indicated. Completeness of DNA methylation was tested by digestion (ThaI), and factor binding was examined in the absence (Ϫ) and presence (ϩ) of nuclear extract from newborn rat FIG. 3. Activities of the proximal ␥D-crystallin promoter and its silencer element during differentiation of explants from 10day-old rats. Transfections were done as described under "Experimental Procedures." The dotted lines represent the activities obtained using newborn rat lens explants as reported previously (7). A, activity of ␥D(Ϫ73/ϩ45)CAT transfected into explants from 10-day-old rats at several stages of bFGF-directed differentiation. Activities are shown relative to that of the maximum level (100%). The bars indicate the standard deviation. B, activity of pBLCAT2␥ (8), containing four copies of the Ϫ85/Ϫ67 silencer region in front of the tk promoter, transfected to explanted lens cells from 10-day-old rats at several stages of differentiation. Activities are shown relative to that of the parental construct pBLCAT2, which was set at 100% (not shown). The bars indicate the standard deviation. The Common Proximal Activator of the ␥-Crystallin Promot-ers-The alignment of the proximal promoters of the ␥-crystallin genes shows that the ␥-box is a well conserved element (Fig.  1) and predicts that all ␥-crystallin promoters, with the possible exception of the ␥B promoter, bind the same factor. Indeed, nuclear factor binding to the ␥D promoter is efficiently competed for by the ␥C and the ␥F promoter (data not shown). The ␥B promoter fragment, however, competed poorly for binding (Fig. 6A). As the ␥B promoter is also the one with the most extended developmental expression, we analyzed the ␥B promoter element in more detail. A promoter fragment containing the ␥B-box yielded two complexes in an EMSA. The lower complex (B2) was competed for by the ␥B but not by a ␥D fragment and thus appeared to be ␥B-specific (Fig. 6B). The upper complex (B1) comigrated with the single complex formed by the equivalent ␥D promoter fragment (not shown) and was also competed for efficiently by this ␥D fragment (Fig. 6B), suggesting that this complex represents the ␥B Ϫ57/Ϫ46 activator (␥B-box) complex. Competition by the ␥B fragment for either the ␥D-box (Fig. 6A) or the ␥B-box complex (Fig. 6B) was very poor, confirming the relatively low affinity of this ␥B promoter element for factor binding. To confirm the conclusion that complex B1 (see Fig. 6B) represents binding to the ␥B Ϫ57/Ϫ46 region, this complex was mapped by in vitro footprinting. As shown in Fig. 6C (left panel), the B1 complex is indeed the result of factor interaction between positions Ϫ55 and Ϫ46. In vitro footprinting of complex B2 showed that the binding site in this complex is the Ϫ46 to Ϫ38 region (Fig. 6C, right panel), directly adjacent to the ␥B-box. The B2 footprint corresponds to a consensus AP-1 site (Fig.  6D). A second AP-1 site is located directly downstream, but no interaction with this site was seen in vitro.
To understand the functional significance of the low affinity binding by the ␥B-box, the ␥D-box was exchanged for the corresponding ␥B element. Mutating the ␥D-box in the ␥D(Ϫ375/ ϩ10)CAT construct successively to the equivalent ␥B sequence led to a gradual decrease in promoter activity (Fig. 7A, left  panel), showing that the ␥B-box is a lesser activator than the ␥D-box. The drop in activity is sharpest between mutant constructs ␥DB1 and ␥DB2, nicely corresponding to the in vitro binding affinity of these mutants; ␥DB1 still efficiently competes with the ␥D sequence, but ␥DB2 no longer does so (Fig.  7B). An even more dramatic effect of mutating the ␥D-box to the ␥B-box is seen when the upstream region is deleted from the Ϫ375/ϩ10 constructs (Fig. 7A, right panel). Deletion to Ϫ73 in the wild-type ␥D promoter has only a slight effect on promoter activity. However, introducing two nucleotide substitutions in the ␥D-box (␥DB2) now results in activity barely above background. Hence, elements in the upstream region contribute to promoter activity, but this effect is only seen experimentally when the ␥-box sequence is less than optimal.
The functional significance of the AP-1 consensus site at Ϫ46/Ϫ38 in the ␥B promoter was tested by mutating this sequence to the corresponding ␥D sequence (␥BD1; Fig. 7C). This mutation severely decreased promoter activity of the ␥B promoter, showing that the AP-1 site acts as an activator (Fig.  7C). This was confirmed by the reciprocal construct, in which the Ϫ46/Ϫ38 ␥B element was introduced at the corresponding site in the ␥D promoter (␥D4); again, this element functioned as an activator, as a 2-fold increase in promoter activity was the result. Finally, we tested the effect of a combination of both the ␥B Ϫ57/Ϫ46 and Ϫ46/Ϫ38 elements in the ␥D promoter. As expected, this construct (␥DB4) had the same low activity as the ␥B-promoter itself.
The ␥D-Box Is a MARE-Kataoka et al. (22) first suggested that the ␥-box might be a Maf recognition element (MARE). This suggestion is supported by the experiments reported by FIG. 4. Nuclear factor from lens or brain binds the ␥D-crystallin silencer at the same site. A, methylation interference footprint analysis of the ␥D-crystallin Ϫ80 region complexed by either lens or brain nuclear factors. ␥D-crystallin Ϫ106/ϩ45 promoter fragment (containing mutation Ϫ46G3 T, thus abolishing factor binding to the ␥-box; see Ref. 7) was 32 P labeled at either end and used to bind nuclear factors from either newborn rat lens or brain. Results using both free (F) and bound (B) DNA are shown. Sites of protein contacts are indicated by brackets. B, summary of footprint analyses shown in A. Residues involved in factor binding are specified by asterisks. Nucleotide sequence is as reported by Den Dunnen et al. (9). C, estimate of the molecular weights of gel-retarded DNA/protein complexes as described by Orchard and May (39). A cloned 32 P-labeled oligonucleotide containing the Ϫ91/Ϫ78 ␥D-crystallin silencer element (␥DS1; see Fig. 5) was bound to either lens or brain nuclear extracts. EMSAs were run together with native standard proteins in a series of gels with increasing polyacrylamide concentration. The relative migration of the DNA/protein complexes (B, complex brain; L, complex lens) and standard proteins (BSA, bovine serum albumin; CA, carbonic anhydrase) were plotted against the polyacrylamide concentration (left). Rf is the migration distance relative to the migration distance of bromphenol blue. The negative slopes of the curves from the standard proteins were then plotted against their molecular weights in a Ferguson plot (39), from which the molecular weight of the lens and brain nuclear complexes was determined (right).
Ogino et al. (40). We have therefore tested whether the factor binding to the ␥D-box also binds the MARE consensus sequence. As shown in Fig. 8, in an EMSA using lens nuclear extract, a ␥D promoter fragment competes efficiently with binding to a consensus MARE. In addition, the mobility of the ␥D-box complex is the same as that of a MARE complex (data not shown), suggesting that the ␥D-box does indeed bind a Maf, at least in vitro.
Two Maf sequences have been reported thus far to be present in the rat lens, Maf-1 and Maf-2 (41,42). Maf-2 is expressed in lens fibers but not in epithelium, whereas Maf-1 is also present in epithelial cells. Cotransfection of expression constructs for Maf-1 or Maf-2 and ␥D(Ϫ73/ϩ10)CAT into differentiating lens explants showed that the ␥D promoter activity was stimulated about 2-fold by Maf-2 but not Maf-1 (data not shown). Hence, the factor binding the ␥D-box might well belong to the Maf family.
The ␥D Ϫ10 Element-The data presented above show that the ␥B-box can function as an activator only in conjunction with the downstream Ϫ46/Ϫ38 element. Yet, the ␥DB3 mutant, which lacks this activator, still retains activity, albeit low (see Fig. 7A). We therefore wondered whether an additional activating element is present in the Ϫ73/ϩ10 ␥D promoter, which, in cooperation with the low affinity ␥DB3-box, drives promoter activity in this mutant. Genomic footprinting of the rat ␥D promoter had revealed a protected site downstream of the TATA box: the GC-rich Ϫ10 region (Ref. 7; see Fig. 1). The nucleotide sequence of this region is unique to the ␥D promoter and absent from the otherwise nearly identical ␥E and ␥F promoters. To test the effect of the ␥D Ϫ10 element, we constructed ␥D/␥F promoter chimeras by replacing the ␥D TATA box and downstream region with the ␥F equivalent, causing the mutations Ϫ21(T3 C), Ϫ18(C3 T), and Ϫ15 to Ϫ12(CGCG3 T-). The latter series of four mutations is situated within the in vivo footprint sequence mentioned above (see Fig. 1). In addition, for practical reasons, the 5Ј noncoding sequence was truncated from ϩ10 to ϩ1.
The activity of the Ϫ375/ϩ1 ␥DF construct was not significantly lower than that of the wild-type ␥D construct when transfected into explanted lens cells (Fig. 9). However, shortening to Ϫ73 caused a drop in activity to about 20% of the corresponding ␥D construct. Introduction of the ␥B-box in the Ϫ73/ϩ1 ␥DF promoter inactivated it, whereas mutation of the G/C-region in construct ␥D4 (see also Fig. 7C) strongly decreased promoter activity. Hence, the ␥D Ϫ10 region acts as an activator. These data further show that the ␥D-box is a relatively poor activator and needs additional elements for full activity. In the ␥D promoter, such elements are located between Ϫ375 and Ϫ73 and around Ϫ10. In the context of our experiments, these elements are redundant. Finally, our results confirm the observation that the ␥B-box is inactive in the absence of other positive elements. However, the positive element does not need to be positioned closely to the ␥B-box, as is the Ϫ46/Ϫ38 element in the ␥B promoter, but can also be located at a distance, as is the Ϫ10 region in the ␥DB3 construct.
Differences between the ␥Band ␥D-Crystallin Silencers-We have shown above that the ␥-crystallin genes share the ␥-box activator. The question arises whether they all have silencers, as predicted from the in vivo mRNA levels (see Ref. 4) and, if so, whether these silencers are common or specific. To examine the presence of silencer elements within the ␥-crystallin promoters, we determined the silencing activity of the upstream regions of the ␥-crystallin promoters (from position Ϫ69) on the heterologous HSV tk promoter. All promoter regions tested repressed activity of the tk promoter when transfected into explanted lens cells, indicating that in all of these promoters, a functional silencing element is present (not shown). Again the ␥B sequence is most divergent and was selected for further analysis.
The Ϫ414/Ϫ69 ␥B fragment, which showed silencing activity in transient transfections, could be deleted to Ϫ110 without loss of silencing activity in lens cells during terminal differentiation (not shown), indicating that the ␥B silencer element must be located between Ϫ110 and Ϫ69. To test whether the ␥B and the ␥D silencing regions bind the same or different factors, the mobility of the ␥B and ␥D complexes was compared (Fig.  10A). Two ␥B complexes were found, both with mobility higher than that of the single ␥D complex, suggesting the formation of different ␥B and ␥D complexes. The non-identity of the complexes was confirmed by competition assays; the ␥D fragment did not compete for the ␥B complexes nor did the ␥B fragment compete for the ␥D complex (Fig. 10A).
In ewborn explants, the ␥D silencer is active only in terminally differentiated fiber cells (Ref. 7; see also Fig. 5B).
To determine whether the ␥B silencer shows the same differentiation-dependent expression, both the ␥B(Ϫ414/Ϫ69)-tkCAT and ␥B(Ϫ110/Ϫ69)-tkCAT fusion genes were transfected to in vitro differentiating lens fiber cells. Like the ␥D element, the ␥B silencer demonstrated differentiation-dependent recognition, as silencing activity was present only during a restricted period of differentiation (Fig. 10B). However, this silencing activity was apparent already between days 4 -7 of differentiation and continued through the terminal stage of differentiation. The Ϫ110/Ϫ69 fragment demonstrated silencing activity only during terminal differentiation, similar to the The ␥-box sequence in a ␥D(Ϫ375/ϩ10)CAT construct was successively mutated to its ␥B equivalent yielding constructs ␥DB1 to ␥DB3 (left panel). ␥B-like sequences are underlined. These constructs and a ␥B(Ϫ414/Ϫ16)CAT construct were transfected into explanted lens cells from newborn rats, following a preculture period of 10 days in the presence of bFGF. Cells were cultured for three more days before harvesting. Promoter activities were determined as described under "Experimental Procedures." Similar experiments were performed with constructs in which the upstream sequences were deleted to Ϫ73 (right panel). All activities shown are relative to that of the parental ␥D(Ϫ375/ϩ10) promoter construct (100%). The bars indicate the standard deviation. B, EMSA showing binding of 32 P-labeled wild-type ␥D(Ϫ73/ϩ10) fragment and lens nuclear extract (ϩ) competed with the ␥D fragment, the equivalent ␥B fragment, or mutant fragments (see A). Binding was in the presence of 50 mM KCl and 50 ng/l poly(dGdC⅐dGdC). Specific competitor DNA was used in 100-fold molar excess. In the first lane, no extract was added (Ϫ). Free (F) and bound (B) probes are indicated. C, functional analysis of the ␥B Ϫ46/Ϫ38 (B2) region. By in vitro mutagenesis, the ␥B Ϫ40 region was exchanged for its ␥D counterpart yielding construct ␥BD1. The reciprocal experiment yielded construct ␥D4. Subsequently, the ␥B-box was introduced in the latter construct, yielding construct ␥DB4. ␥B-like sequences are underlined. Arrows point to mutated nucleotides. Constructs were transfected into explanted lens cells from newborn rats, which had been precultured for 10 days in the presence of bFGF. After three additional days of culture, the cells were harvested, and promoter activities were assessed as described under "Experimental Procedures." Activities are shown relative to the wild-type ␥D construct (100%). The bars indicate the standard deviation.
␥D silencer. Although the extent of silencing by the Ϫ110/Ϫ69 fragment is similar to that of the larger Ϫ414/Ϫ69 fragment during the course of differentiation, surprisingly, in early differentiation it strongly activated the tk promoter. Hence the Ϫ414 to Ϫ69 region of the ␥B-crystallin gene must contain additional enhancers and silencing elements. DISCUSSION A simple mechanism for developmental regulation of the promoter activity of the ␥-crystallin genes would be that the level of activity is determined by the affinity of a common activating factor for the proximal activator, the ␥-box. One would then predict that a ␥-crystallin promoter that is shut down early in development (e.g. ␥E, ␥F) would have a low affinity ␥-box, whereas a gene of which expression continues until later in development would have a high affinity ␥-box. Our data clearly show that this hypothesis is not correct; we find that the ␥-crystallin gene with the most sustained expression during development, the ␥B gene, has the ␥-box with the lowest affinity.
Comparison of the ␥B-box sequence with that of the ␥D-box shows that in the ␥B-box the C at Ϫ54, which shows a protein contact in vivo (7), has been replaced by a T. This suggests that it is the 5Ј-half of the binding site in the ␥B-box that is responsible for the low affinity. However, our results show that even such a scrambled site is sufficient to target the corresponding cognate factors to the promoter, providing that additional activating elements are present. The additional activating element can be either a closely linked AP-1 site, as in the ␥B promoter, or the more distant up-and downstream elements of the ␥D promoter. Hence, there is no constraint on either the nature or the distance of the additional activating element. The ␥-box resembles a MARE in sequence, and binding of a cognate factor in lens extracts is competed by a consensus MARE sequence. Furthermore, the ␥D promoter is slightly activated by cotransfection of an expression vector for Maf-2. Hence, the in vivo activator of the ␥-crystallin promoter might well belong to the Maf family of transcription factors. Interaction of the mouse ␥F-box with a (chicken) zinc finger protein has also been reported (11). However, this protein acts as a transcriptional repressor rather than as an activator.
The Maf family is a diverse one with small and large members (for recent reviews, see Refs. [43][44][45]. The large members, which include the founding member of this family, v-Maf, have an N-terminal acidic activation domain. The small members, such as MafF, MafG, and MafK, lack this activation domain and can activate transcription only as heterodimers with, for example, a large Maf member or a member of the AP-1 family (e.g. see Refs. 22,23,45). Two large Maf family members have thus far been shown to be present in the rat lens. Maf-1 is the rat homologue of the mouse MafB, and Maf-2 may be the homologue of the chicken c-Maf (41,42). Maf-1 mRNA is primarily found in the lens epithelial layer; Maf-2 mRNA is prominent in the fiber cell mass, the site of expression of the ␥-crystallin genes. Expression of Maf-2 is not limited to the lens, since Maf-2, as well as Maf-1, is also found in many other tissues of the body such as kidney, spleen, and liver. Hence, a role for Maf-2 in the expression of the ␥-crystallin promoters is seemingly at odds with the lens specificity of these promoters in transgenic mice (13,14) or even in transgenic Xenopus laevis (46). However, Maf-2 could partner a lens-specific protein, and the role of Maf-2 in directing lens-specific expression could be analogous to that of MafK in erythroid-specific transcription; MafK acts as the partner of the erythroid-specific transcription factor NF-E2 (47). Alternatively, the "true" activator of the ␥-crystallin promoters could be another Maf protein. A possible candidate would be the rat homologue of L-Maf, a lens-specific FIG. 8. The ␥D-box resembles a MARE. EMSA showing binding of a double-stranded MARE consensus oligonucleotide (TGCTGACTCAG) and lens nuclear extract from newborn rats (ϩ) competed for with a ␥D(Ϫ73/ϩ10) fragment. As control, competition with the MARE sequence is also shown. Binding was in the presence of 50 mM KCl and 50 ng/l poly(dGdC⅐dGdC). Specific competitor DNA was added in 10-and 100-fold molar excess. In the first lane, no extract was added (Ϫ). Free (F) and bound (B) probes are indicated.
FIG. 9. The ␥D-crystallin ؊10 region acts as an activator. In the ␥D-crystallin constructs indicated (see also Fig. 7), the TATA box and downstream region were exchanged for the ␥F equivalent, and the Ϫ375/Ϫ73 region was deleted. The ␥D-box is denoted by ␥-box, in ␥DB3 the ␥D-box has been mutated to the ␥Bbox sequence, B2 represents the AP-1 consensus sequence as found in the ␥B promoter. Constructs were transfected into newborn rat lens explants, which had been precultured for 10 days in the presence of bFGF. After three additional days of culture, the cells were harvested, and promoter activities were assessed as described under "Experimental Procedures." Activities are shown relative to the wild-type ␥D(Ϫ375/ϩ10) construct (100%). The bars indicate the standard deviation. member of the Maf family found in the chicken lens (40). 3 However, it is not yet known whether such a rat homologue indeed exists. Clearly, further experiments are required to elucidate the role of Maf proteins in the regulation of ␥-crystallin gene expression.
AP-1 elements are targets of the Fos and Jun family of transcription factors, well known for their role in transmitting growth factor signals to the transcriptional apparatus. The involvement of an AP-1 element in the activity of the ␥B promoter suggests that the activity of this promoter is subject to regulation by extracellular factors. This could play a role in the developmental regulation of the activity of this gene and could further explain our rather puzzling observation that the level of ␥B mRNA reached during in vitro differentiation is significantly lower than that found in vivo. 2 Apparently, bFGF is not capable of providing the proper signals for maximal activation of the ␥B promoter during in vitro differentiation.
The ␥-crystallin promoters contain an invariant sequence (CCCTTTTGTG) located Ϫ35 base pairs upstream from the TATA box (Ϫ73 to Ϫ63 in the ␥D promoter). The TTTG region in this sequence has been shown to be a binding site for Sox-2, a member of the Sry family of transcription factors (Ref. 48; see Fig. 11). The CCC are contacted by a factor in vivo, as they are detected on the in vivo footprint of the ␥D promoter (7). In that paper, the assumption was made that these CCC formed part of the silencer element. However, we show here that the silencer is located further upstream and contacts the bases Ϫ90 to Ϫ78. The close proximity of the CCC footprint at Ϫ73/Ϫ71 to the Sox-2 target site suggests that this footprint belongs to a factor that forms a heterodimeric complex with Sox (note that Sox binds in the minor groove, whereas in vivo DMS footprinting detects only major groove G contacts). We have not studied the effect of this site on ␥-crystallin promoter activity here as the sequence is invariant. We have previously shown that deletion of Ϫ77/Ϫ71 in the ␥F promoter caused a 60% drop in activity (35).
Together, our studies show that there is a plethora of positive and negative factor interactions in the proximal ␥-crystallin promoter (see Fig. 11). During development, the balance between these factors shifts toward repression. At least for the ␥D-crystallin promoter, it is the developmental change in the pattern of expression of the silencing factor that is the primary cause of promoter repression. Lack of demethylation is probably a secondary cause. However, the rate of promoter demethylation progressively decreases during development, and at even later developmental stages the rate of promoter demethylation may well be too slow to allow transcriptional activation, even if the cognate transactivating factors are present. Note that demethylation of the ␥D promoter region cannot be passive, i.e. due to lack of maintenance methylation following DNA replication, but must be active as there is no cell division coincident with promoter demethylation in differentiating lens explants. 2 32 P-labeled fragments of the ␥B-(Ϫ110/-69, (BS) and ␥D-crystallin (Ϫ93/Ϫ76, ␥DS1) promoters were incubated with rat lens nuclear extract and electrophoresed through a native polyacrylamide gel (ϩ). Fragments were bound in the presence of 10 mM KCl and either 50 ng/l poly(dGdC⅐dGdC) (in the case of the ␥B probe) or 50 ng/l poly(dIdC⅐dIdC) (in the case of the ␥D probe). Specific competitor DNA was added in a 100-fold molar excess. In the first and fifth lanes, no extract was added (Ϫ). Free (F) and bound (B) probes are indicated. See "Experimental Procedures" for details. B, functional analysis of the ␥B-crystallin silencer region. The ␥B(Ϫ414/Ϫ69)-tkCAT and ␥B(Ϫ110/ Ϫ69)-tkCAT constructs were transfected to newborn rat lens explants, precultured with bFGF for the time as indicated. The cells were harvested 3 days later, and promoter activities were determined as described under "Experimental Procedures." Activities are shown relative to that of the tk promoter (pBLCAT2; 100%). The bars indicate the standard deviation. sequence elements that signal ␥D promoter demethylation are unknown. Mapping these elements has thus far been precluded by the low transfection efficiency of lens explants at early stages of differentiation. Our efforts are now directed at overcoming this practical problem.