Expression in non-lens tissues of an enzyme activity related to the ‘lensspecific’ protein, δ crystallin

It has been known for more than a decade that several non-lens tissues in early chick embryos express low levels of δcrystallin, a structural protein usually considered specific to the embryonic lens fibres (Clayton et al. 1979; Agata et al. 1983; Kodama and Eguchi, 1982; Nomura, 1982; Jeanny et al. 1985; reviewed by Clayton et al. 1986; de Pomerai, 1988). The reason for such ‘ectopic’ δexpression remains obscure, although immunocytochemical studies (Linser and Irvin, 1987) suggest that this feature is confined to particular cells, e.g. a subset of neuroretinal glial cells around the optic nerve boundary in chick embryos. It is also known that some (but not all) of these δ-expressing non-lens tissues can transdifferentiate in long-term cell culture under appropriate conditions, giving rise to lentoid bodies composed of lens-fibre-like cells rich in δ and other crystallins. This category of tissues includes embryonic tapetum (Eguchi and Okada, 1973), neural retina (NR; Okada et al. 1975; de Pomerai, 1988) and early embryonic brain (Nomura, 1982), as well as quail pineal gland (Watanabe et al.

1985). However, both early embryonic limb bud (Kodama and Eguchi, 1982) and later embryonic brain (Takagi, 1986) show enhanced δ crystallin production during culture without concomitant formation of lentoidal cells; thus high-level δ expression can be divorced from the lens-fibre phenotype. Still other embryonic tissues express δRNA (and possibly protein) in some cells, but later this fades out both in vivo and during culture, as in the case of heart and liver (Jeanny et al. 1985), or adenohypophysis (Ueda and Okada, 1986). While ectopic δ expression by tissues in the first category might conceivably predispose the δ-positive cells towards transdifferentiation into lens in culture, this clearly fails to explain why such cells express δin the first place, since no lens-like cells appear at any stage during the differentiation of these tissues in vivo. No convincing explanation for this phenomenon has been forthcoming, apart from imprecise suggestions of ‘leaky’ transcriptional control for the δ gene(s) and/or some hitherto undefined function for low levels of δ crystallin in non-lens tissues.

There are two closely related and linked δcrystallin genes in the chicken genome, designated δ1 and δ2. Both have 17 exons, most of which are 90 – 100% identical in sequence between δ1 and δ2; even the introns show extensive sequence similarities, and the two genes differ mainly at their 5’ ends (Nickerson et al. 1985, 1986). Nevertheless, there is evidence that <51 is expressed at approximately 100-fold higher levels than δ2 in lens tissue (Parker et al. 1988), and studies with promoter/CAT fusion genes suggest a significant difference in δ1 versus δ2 promoter activity (Borras et al. 1985). Notably, the δ1 crystallin gene also contains an enhancer in its third intron, which confers abundant lens-specific expression (Hayashi et al. 1987). Studying any differential regulation of these two δgenes in nonlens tissues is made difficult by the extreme similarity of the δ gene-products as well as by their low prevalence. Recently, computer sequence searches have suggested an unexpected relationship between chick δ crystallins and the mammalian urea-cycle enzyme ASL (Piati-gorsky et al. 1988). Current evidence points to δ2 as the chicken ASL gene, while δ1 appears to be a variant copy encoding the major structural protein of embryonic lens fibres. Thus one reason for low-level iectopic’ δ expression might be a metabolic requirement for ASL activity (the δ 2 product) in certain tissues or cells.

The present paper reports the results of two studies related to this point. The first shows that the chromatin conformation of the δ locus changes during NR transdifferentiation into lens in vitro. Late (fully transdifferentiated) NR cultures resemble lens both in terms of the DNAasel-sensitivity of the δ locus and the level of δ gene transcription detectable in vitro’, this last is much higher than in early embryonic NR tissue. The second study shows that ASL activity is correlated with the presence of a 48 – 50 × 10³M_r protein recognised by a polyclonal anti-δ -crystallin antibody in a range of tissue samples from chick, mouse and amphibian (Ranaf, a similar protein is recognised by this antibody in a commercial bovine ASL preparation. In the cases of chick lens and transdifferentiated NR cultures, high levels of immunodetectable δ protein appear to confer relatively low ASL activities. Tentatively, we suggest that abundant lens-fibre-specific expression of 61 protein might dilute the contribution of δ /ASL in these instances. If so, the known cases of ‘ectopic’ <5 expression might involve a higher proportion of δ 2/ASL (as compared to δ 1) products than is the case in lens cells.

Fertile chicken eggs from a single commercial broiler strain were obtained from G. W. Padley Ltd (Grantham, UK). All radiochemicals were from Amersham International pic, medium components and sera from GIBCO-Europe Ltd, tissue culture disposables from Nunc or Sterilin, and most chemicals from Sigma Ltd (including ASL and immunological reagents). The full-length δ crystallin cDNA probe used in this study (p δCr17) was a generous gift from Dr J. Piatigorsky (NIH, Bethesda, Maryland, USA). The polyclonal anti-6 crystallin antibody was raised in rabbits; after absorption with total headless-chick-embryo soluble proteins, it was monospecific by immunodiffusion and immunoblotting criteria (de Pomerai et al. 1984; the same antibody was used by Linser and Irvin, 1987).

Eyes were dissected from chick embryos after 4 or more days of development and separated carefully into lens+vitreous and neural retina (NR)+tapetum fractions. Lenses from 17-day embryos were manually dissected to give central fibre masses (LF) and epithelia plus peripheral fibres (LE). NR was separated from tapetum by gentle shaking after incubation for 10 min in Ca²⁺- and Mg²⁺-free saline (CMF; 137 min NaCl/5.3mM KC1/5.6ITIM glucose/3.1 HIM Na₂HPO₄/0.4mM KH2PO4, pH7.0). For cell culture purposes, NR from 7- or 8-day eyes was trypsin dissociated and the cells cultured for up to 40 days in permissive FH medium (Eagle’s MEM with Earle’s salts plus 26mM NaHCO₃, 2mM L-glutamine, lOOi.u.ml^-1 penicillin, 100 pg ml^-1 streptomycin, 5% horse serum and 5% foetal calf serum; de Pomerai and Gali, 1982), at a density of 2x10⁷ cells per 6cm culture dish. Such cultures were maintained at 37°C in a humid atmosphere of 5 % CO₂:95 % air, the medium being changed every 2 – 3 days.

These procedures were carried out as described previously (Carr and de Pomerai, 1985), except that VRCs were present at 5mM during RNA isolation to minimise degradation by RNAases. Poly(A)⁺ RNA was isolated using Hybond messenger-affinity paper according to the manufacturer’s recommendations (Amersham International pic). Size calibrations for the Southern blots used a parallel restriction digest of λ DNA, stained with ethidium bromide and visualised under UV.

Protein separations using SDS – PAGE, and subsequent Western blotting onto nitrocellulose membranes (Schleicher & Schuell), were performed as described previously (Carr and de Pomerai, 1985). After blocking the membrane blots overnight in Tris-buffered saline (TBS; 110 mM NaCl/10mM Tris-HCl, pH7.5) containing 3% (w/v) BSA and 5% (v/v) horse serum, blots were incubated first for 6h in a 1:2000 dilution in TBS of our anti-6-crystallin antibody, then second (after extensive washing in TBS with and without 3% BSA) for 2h in a similar dilution of peroxidase-linked anti-rabbit IgG (Sigma). After further washing, blots were stained using 4-chloro-l-naphthol and hydrogen peroxide.

Immunological quantitation of δ crystallin levels by means of haemagglutination inhibition assays (4 to 6 replica assays per sample) was also carried out as described previously (de Pomerai and Gali, 1981, 1982). Immunoprecipitations were performed using 0.5 ml of a CMF extract prepared from 80 7-day NRs plus 30 μl of anti-δ antibody. Both extract and antibody were centrifuged (11000g for 10 min) prior to mixing. After 4 h at 4°C, the mixture was recentrifuged (same conditions) and the pellet retained as immunoprecipitate 1. The supernatant was then mixed with 0.2 ml of protein A-sepharose CL 4B (Ellis et al. 1987) and recentrifuged to bring down remaining immune complexes (immunoprecipitate 2). Both immunoprecipitates were washed 6 × with 1ml of TBS, and immune complexes were dissociated with 80 μl of TBS containing 0.1% SDS; 10 μl samples were run on 5 parallel lanes of an SDS gel as above. A covalent immunoselection reagent was prepared by linking the antibody onto CNBr-activated Sepharose 4B (Sigma) according to the manufacturer’s recommendations. Preliminary experiments (not presented) showed that this reagent would selectively bind δ crystallin, and that δ was the only major protein recovered from the reagent pellet after washing with TBS and dissociation with 0.1% SDS/TBS. The reagent (0.2 ml packed gel) was incubated for 4h at 4 °C with 0.5 ml of 7-day NR extract (as above); after washing 6 × with 1ml of TBS, immune complexes were dissociated with 0.2 ml of 0.1% SDS/TBS, centrifuged (11000g for 5 min), and the supernatant dialysed for 6h at room temperature against 3 × 500ml changes of TBS. Aliquots of this were subjected to gel electrophoresis and blotting, or assayed for ô content (as above) or ASL activity (below).

ASL activities were determined by the colorimetric method of Campanini et al. (1970), except that 5% NaOCl was used in the final stage to intensify the colour. 4 to 6 replicate assays were performed on aliquots of CMF extracts of soluble proteins prepared from the tissues indicated. These assays were calibrated using purified bovine ASL (Sigma) of known specific activity. Protein contents were determined for all tissue extracts by the method of Lowry et al. (1951). Mean ASL activities are expressed in units per mg soluble protein, using the unit definition of Sigma Ltd. Following protein transfer from SDS gels, transverse strips of nitrocellulose membrane (across 4 identical gel lanes) were also assayed for ASL activity. Three such strips (each 1 cm wide) were tested, with their centres respectively (i) within the δ band (determined by probing a fifth parallel lane with anti-ô, as described above), (ii) 1.5 cm above and (iii) 1.5 cm below this position.

Lens and NR nuclei were isolated by the RSB method of Rymo et al. (1974), using ice-cold RSB isolation buffer (10 mM Tris-HCl pH 7.4/10 mM NaCl/2mM EDTA/0.5mM EGTA/ 0.15mM spermine/0.5 mM spermidine/1 mM phenylmethylsul-phonyl fluoride) supplemented with 0.2 % (v/v) Nonidet P-40 (Weintraub and Groudine, 1976) to effect cell lysis. Nuclei were passed through a 10 μ m nylon mesh (H. Simon, Stockport, UK) to remove unlysed cells and then pelleted at 11000g for 5 min at 4°C. Prior to lysis in the above buffer, NR cultures were partially dissociated using 0.025 % trypsin/ 0.02% collagenase in CMF for 40 min at 37°C. Gentle homogenisation was used to assist in the release of nuclei from both 14-day embryonic lenses and NR cultures. Nuclei of all types were then washed in digestion buffer (10 mM Tris-HCl pH 7.4/10 mM NaCl/3mM MgCl₂/0.1mM CaCl₂) prior to digestion with DNAasel and/or EcoRl (see below). For in vitro transcription purposes, nuclei were isolated from fresh 5-day NR tissue by the method of Tata (1974), from late NR cultures by the procedure of Mory and Gefter (1977) and from lenses by the RSB method (above). In each case, the method indicated gave optimal yields of nuclei and/or maximal rates of transcription in vitro (data not shown).

2 × 10⁷ nuclei prepared by the RSB method were resuspended in 0.5 ml of digestion buffer (above) and treated with micrococcal nuclease or DNAasel at 0.5 to 10 Kunitz units per ml (see legend to Fig. 1) for 10 min at 37°C. The reaction was terminated by adding EDTA to 10mM, SDS to 0.2% (w/v) and proteinase K to 200 μ g ml^-1. After 3h at 37°C,

DNA was extracted by the method of Gross-Bellard et al. (1973). The DNA was restricted using EcoRl (Northumbria Biologicals, UK) as recommended by the manufacturer. Agarose gel electrophoresis, Southern blotting and hybridisation with a nick-translated ³²P-labelled probe (in this case, p<5Crl7) were all carried out as described by Carr and de Pomerai (1985).

5 × 10⁷ nuclei prepared by the RSB method were resuspended in 100 μ l of 50mM Tris-HCl pH 7.8/5 mM MgCl₂ β -mercaptoethanol/10 μ gml^-1 bovine serum albumin. Nick translation was carried out for 30 min as described by Gazit et al. (1982), using very low concentrations of DNAasel and 9.3 MBq each of ³²P-labelled dCTP and diTP (specific activity HOTBqmmol^-1). DNA was extracted as described above, then after mild mechanical shearing, about 5 × 10⁶ctsmin^-1 from each assay (Hutchison and Weintraub, 1985) was hybridised as above to a strip of Hybond N membrane (Amersham International pic) carrying unlabelled size-separated PstI fragments of the p δ Crl7 plasmid (cf. Groudine et al. 1981), which had been bonded to the membrane by UV treatment. The Southern blotting transfer time was increased to 72 h to facilitate binding of the small 80/85 bp Pstl fragments.

Nuclei prepared from 5-day fresh NR (250-1000 μ g DNA) or 14-day lens (20 μ g DNA) or 40-day NR cultures (50 μ g DNA) were assayed as described by Groudine et al. (1981), using 370 kBq per assay of o’-³²PO₄-labelled UTP in a total volume of 100 μl (containing 30% glycerol/2.5 mM dithiothreitol/ ImM MgCl₂/70mM KCl/0.5mM ATP/0.25mM each of CTP and GTP). Preliminary experiments showed that RNA yields are greatly increased by additionally including RNAasin (human placental ribonuclease inhibitor from BDH; 30 units/ assay) and VRCs (ImM) during the assay. Conversely, inclusion of a-amanitin (2 μgml^-1) inhibited transcription by at least 70%. Transcription continues linearly for at least 2h under these conditions (data not shown). After 3h at 30°C, transcription was terminated by adding 100units/assay of RNAase-free DNAasel plus VRCs to 5 mM final; this mixture was incubated overnight at 35°C to digest the DNA completely (shorter incubations reduced the yield of RNA). SDS was then added to 1 % (w/v), EDTA to 5 mM, Tris-HCl (pH7.8) to 10mM, VRCs to 20mM and proteinase K to 100pg ml^-1 in a final volume of 0.5 ml. After deproteinisation for 1h at 42 °C, RNA was phenol-chloroform extracted and precipitated overnight with propanol (50% v/v) in the presence of sodium acetate (0.15 M) and tRNA carrier (100 μg ml^-1). The labelled RNA (usually >10⁵cts min^-1 per assay), was then hybridised to pieces of Hybond N membrane carrying 1μg dots of denatured unlabelled p δ Cr17 DNA (bonded to the membrane by UV treatment as above; hybridisation conditions as described by Groudine et al. 1981). After extensive washing to remove unbound RNA, the dots were dried, autoradiographed and counted.

Fig. 1A shows the locations of restriction sites recognised by EcoRl within the chick <5 locus, as determined by computer searching the published δ 1 and δ 2 gene sequences (Nickerson et al. 1985, 1986). In our study, all of the major EcoRl fragments (11.0, 7.6,5.3, 3.9 and 3.0 δbp) are picked up by the full-length <51 cDNA probe (pδCrl7), although the small 0.7kbp fragment does not usually transfer efficiently, and the 5.3kbp fragment tends to give a rather weak signal because of sequence disparities between the 5’ exons of δ1 and δ2. Restriction fragment length polymorphisms, reported previously for the δ locus (de Pomerai and Carr, 1985; Nickerson et al. 1985, 1986), do not appear to be a significant factor affecting our results. In any case, all DNA samples are derived from pooled material contributed by 20 or more chick embryos (as fresh or cultured tissues). Nuclei prepared by all methods appear morphologically intact, and give standard nucleosomal ladders following micrococcal nuclease digestion (e.g. Fig. IB), suggesting that there is little disruption of the basic chromatin organisation during nuclear isolation.

Fig. 1C shows that pretreatment of nuclei from fresh 8-day NR or 15-day cultures with high levels (5–10 units ml^-1) of DNAasel does not alter the subsequent pattern of EcoRI digestion in the – locus. Thus the ô locus is DNAasel-resistant in the majority of nuclei in fresh 8-day NR and in early 15-day cultures. By contrast, DNAasel pretreatment at 1.0 unit per ml almost abolishes the EcoRI restriction pattern when using nuclei from 20 or 35 day (transdifferentiated) cultures of 8-day NR cells (Fig. 1C and ID). Examination of the gels under UV prior to blotting and hybridisation confirmed that all lanes contained equivalent loadings of DNA and that the average fragment size was not altered by DNAasel treatment even at 10unitsml^-1. Thus the δ locus becomes DNAasel-sensitive between 15 and 20 days of culture in the majority of nuclei (Fig. 1C). These results are confirmed by the in vitro nick translation data shown in Fig. 2. Extremely low levels of DNAasel were used to introduce nicks selectively into DNAasel-sensitive regions of chromatin, which were then labelled by nick translation. Labelled nuclear DNA was extracted and hybridised to Southern blot strips carrying unlabelled size-separated PstI fragments of the p δCr17 plasmid (see Fig. 2A). A control experiment, hybridising labelled p δl7 to its own PstI fragments, gives a strong signal from the 500 and 450 bp fragments, but a weaker signal from the large 1150 bp (mainly plasmid) and smaller 80/85bp fragments (Fig. 2B, lane 1). Nick-translated nuclear DNAs from both lens and 40-day (transdifferentiated) cultures of NR cells bind to the 500 and 450 bp fragments (Fig. 2B, lanes 2 and 4), whereas no signal is detectable using nuclei from earlier 10-day cultures of these cells (Fig. 2B, lane 3). Thus the δlocus is DNAasel-resistant in the majority of nuclei from early NR cultures, but is DNAasel-sensitive in most nuclei from transdifferentiated cultures as well as in those from lens.

Nevertheless, one or both of the δ genes must be active in at least a small proportion of the nuclei in fresh embryonic NR tissue. As shown in Fig. 3A, low levels of δRNA are detectable in 4-, 7- and 11-day fresh NR (lanes 4, 5 and 6), although the amounts present are much less than 1 % of those in newly hatched chick lens (lane 3); δ transcripts are very much rarer in adult hen liver and virtually absent from oviduct (lanes 1 and 2). These data are in full agreement with previous reports (Clayton et al. 1979; Agata et al. 1983). Overexposure of these autoradiographs (not shown) confirms that traces of δ RNA are present in adult liver and possibly in oviduct; although birds are uricotelic rather than ureotelic, it is possible that some liver cells do express ASL (δ2) at low levels. Using nuclei from 5-day embryonic NR, we detect very low levels of labelled in vitro transcripts hybridising with the δ probe (Fig. 3B). The levels of such δ transcripts are (i) roughly dependent on the concentration of nuclei (shown separately for two different nuclear preparations), (ii) significantly reduced when RNAasin is absent during the assay (implying rapid RNA processing), and (iii) largely abolished in the presence of 2 μgml^-1α-amanitin. Thus very small amounts of <5 RNAs are detectable among the in vitro transcription products synthesised by endogenous RNA polymerase II in isolated nuclei from 5-day embryonic NR. Preliminary dot-blot experiments (Fig. 3C) suggest that the levels of αtranscripts synthesised by 5-day NR nuclei (part c) are much lower than those obtainable from 10-fold smaller numbers of transdifferentiated NR culture nuclei (part b) or from 25-fold fewer lens nuclei (part a). In summary, the α locus must be active at low levels in at least some nuclei in fresh embryonic NR tissue. However, widespread activation of this locus in surviving cultured cells is implied by the change from overall DNAase 1-resistance to sensitivity during NR transdifferentiation into lens. Attempts to distinguish α1 from α2 transcripts (cf. Parker et al. 1988) in fresh or cultured NR cells have so far been unsuccessful in our hands.

In Fig. 4 we show the results of an immunoblotting survey using a polyclonal but monospecific antibody raised in rabbits against chick δcrystallin. The only polypeptide bands reacting with this anti δantibody on Western blots are a major species of about 48 × ICr,M_r and a minor component of about 50 ×10³A/_r (cf. Reszelbach et al. 1977; Shinohara and Piatigorsky, 1977; Fig. 4, lane 1). These proteins are probably both derived from δ1 mRNAs by cotranslational modification of the N-terminal regions of the nascent δ1 polypeptide (Wawrousek and Piatigorsky, 1987). The presence of several δ polypeptides is also shown by twodimensional gel analysis of chick lens proteins (Thomson et al. 1978; de Pomerai et al. 1984); it is not clear which of these δ species derive from δ 2. In addition, our antibody picks out a few much fainter bands of higher molecular weights in lanes overloaded with lens proteins (Fig. 4, lanes 5 and 6); we speculate that these may represent undissociated oligomers of δ crystallin. The sensitivity of the antibody detection system is at least 4-fold greater than that of Coomassie blue staining for lens protein samples (Fig. 4, lanes 1 to 3). Immunodetectable δ crystallin is clearly present in 7-day fresh NR tissue, though it is barely detectable among 11- or 14-day fresh NR proteins at comparable loadings (Fig. 4, lanes 7 to 9). However, δ crystallin is a major protein constituent of transdifferentiated 40-day NR cultures (Fig. 4, lane 10). In addition, we have detected protein bands of 48 – 50 × 10³ A7_r which crossreact with our anti-δ antibody in the following non-chick samples: (i) amphibian (Rana pipiens’) lens and liver (Fig. 4, lanes 11 and 12); (ii) mouse liver (lane 13); and (iii) commercial bovine argininosuccinate lyase (ASL; lane 4). Provisionally, we identify the mouse and Rana δ -related bands as ASL homologues, since δ2 is the chick ASL gene (Piatigorsky et al. 1988), but neither mammals nor amphibia have δ -type crystallins as major lens proteins. Table 1 shows that tissue samples containing δ -related protein bands (above) also have significant ASL activity. The haemagglutination inhibition assays used to quantitate chick δ -crystallin levels (de Pomerai and Gali, 1981, 1982) did not give clear endpoints with the mouse or Rana samples, even though the same antibody recognises a δ-related protein on the Western blots (Fig. 4). In most cases, Table 1 indicates that relatively low levels (<0.1 %) of immunodetectable δ are correlated with ASL activities in the range 0.04 to 0.64 × 10³ units mg^- * protein. However, three further samples show much higher δ contents (9 to 60%) corresponding with ASL activities of only 0.97 to 5.2 ×10⁻³unitsmg^-1 protein. These three are (i) 17-day central lens fibres (LF), (ii) 17-day lens epithelium plus peripheral fibres (LE) and (iii) transdifferentiated 40-day NR cultures. Since all three contain abundant lens-fibre-like cells, they are likely to express high levels of δ1 protein which may dilute the δ2/ASL product (cf. Parker et al. 1988).

Immunoprecipitation experiments, followed by SDS – PAGE/Westem blotting and ASL assays on strips of membrane, confirm that the ASL activity in 7-day NR tissue is largely conferred by an immune-selected protein band in the δ region (Table 2). Both immunoprecipitates (1 and 2, see Methods; Fig. 4, lanes 18 and 19) contain significant ASL activity in the δ-position strip but not in the other two, whereas no such enrichment is found in the supernatant fraction (Fig. 4, lane 20) and much less in the original 7-day NR extract (Fig. 4, lane 17). Finally, the anti-δ antibody/ CNBr-Sepharose reagent selects a 48 –50 × 1(‘P,<W_r <5-related band from among the proteins present in a 7-day NR tissue extract (Fig. 4, lanes 14 – 16); notably the selected fraction is greatly enriched in ASL activity (last line of Table 1). This evidence strongly suggests that the major source of ASL activity in 7-day NR is a 48 – 50 × 10³M_r protein recognised by our anti antibody -presumably the δ 2 protein.

Our results demonstrate a change in the chromatin conformation of the δ locus, from predominantly DNAase 1-resistant to mainly DNAasel-sensitive, in the nuclei of cultured NR cells during transdifferentiation into lens (Figs 1 and 2). This apparently precedes the activation of δ expression in the proportion of cells (usually 20 – 30%) that eventually become lens-fibre-like during lentoidogenesis in these cultures. However, more than half of the cells in late NR cultures do not express immunodetectable δ crystallin (Okada et al. 1975). We find no evidence for cultured NR cells in which the δ locus remains DNAasel-resistant (which would result in a weaker but continuing signal for the EcoRI pattern in DNAasel-treated cultures, not seen in Fig. ID). It follows that this locus must be DNAasel-sensitive in the nuclei of many cells that are non-lentoidal and do not express the δ protein. This may relate to the fact that δ transcript expression (sometimes confined to the nuclei) is very widespread in late NR cultures under conditions permissive for transdifferentiation (Carr and de Pomerai, 1985). Thus DNAasel-sensitivity and even transcription of the δ locus are not diagnostic of the lens-fibre phenotype, but are general features of late NR cultures, extending to cells of glial or tapetai morphologies, which usually outnumber the lentoidal cells. Preliminary time-course studies examining the appearance of DNAasel-sensitivity in the δ locus during NR culture suggest that this locus remains resistant in 10- or 15-day cultures but becomes sensitive by about 20 days (Fig. 1C, lanes 3 and 4). This changeover considerably precedes the onset of δ synthesis and the appearance of lentoids, but coincides roughly with the time of commitment to transdifferentiation as determined by transfers between permissive and non-permissive media (Gali and de Pomerai, 1982). Changes in the methylation status of the δ locus have previously been described during lens differentiation (hypomethylation in postmitotic lens fibres; Sullivan and Grainger, 1986), but no such changes have been detected during NR transdifferentiation (Errington et al. 1983). In any casé the changes in lens do not coincide with the onset of δ gene transcription.

δ-crystallin transcripts and sometimes protein have been detected in a variety of chick embryonic tissues, including neural retina (Clayton et al. 1979; Agata et al. 1983; Bower et al. 1983; de Pomerai, 1988; Fig. 3A). Our data also show, for the first time, that one or both δ genes are transcribed at low levels during in vitro transcription in isolated nuclei from fresh 5-day embryonic NR (Fig. 3B). Much higher levels of δ transcription are found in nuclei from lens or transdifferentiated NR cultures (Fig. 3C). This disparity may involve both the rate of δ transcription and/or the number of cells involved. Immunocytochemical and in situ hybridisation studies (Linser and Irvin, 1987; Bower et al. 1983) suggest that only a minority of cells in fresh NR tissue actually engage in δ synthesis. A more accurate quantitative comparison of δ transcription in these various types of nuclei will require a much more sensitive small-scale in vitro transcription system, such as that described recently for lens by Zelenka et al. (1989). Nevertheless, it is clear that late NR cultures resemble lens both in terms of the level of δ transcription (Fig. 3C) and in the DNAasel-sensitivity of the δ locus (Fig. 2B). This is not to imply that the chromatin structure and/or transcriptional regulation of the δ locus need be identical between lens and late NR cultures. Indeed, preliminary comparisons between these tissues suggest that the locations of some DNAasel-hypersensitive sites in the δlocus may be shared but others different (de Pomerai, 1988). One obvious question that arises from these studies concerns the relative contributions of δ1 and δ2 gene products in lens or lentoid as compared to non-lens cells. We have tried various approaches to make this distinction in fresh NR tissue, but have so far failed, due partly to the low levels of such transcripts and partly to their close sequence similarity (cf. Parker et al. 1988).

The recent identification of δ2 as the chicken ASL gene (Piatigorsky et al. 1988) has permitted an alternative, if indirect, approach to this question. We find good correlation between ASL activity (Table 1) and the presence of a 48 – 50 × 10³M_r polypeptide recognised by our polyclonal anti-δ antibody (Fig. 4). Immune selection studies confirm that this protein is indeed the major source of ASL activity in 7-day chick embryo NR (Tables 1 and 2; Fig. 4). We note that the amounts of immunodetectable δ (on Western blots) seem to decrease sharply between 7- and 11-day NR samples, whereas the levels of δ RNA appear similar in both (compare Fig. 4, lanes 7 and 8, with Fig. 3A, lanes 4 and 5). Since the ASL activities (Table 1) suggest only a slight drop between these two stages, it is not clear whether these differences are significant. Chick lens samples (LE and LF) contain at least 400-fold higher levels of total δ crystallin as compared to 7-day NR, but the corresponding ASL activities are only about 10-fold higher. Provisionally, we attribute this difference to the fact that some 99% of the δ mRNA in lens is derived from the δ1 gene and only 1 % from the δ 2/ASL gene (Parker et al. 1988; Piatigorsky et al. 1988). Since transdifferentiated NR cultures also contain high levels of Ô but relatively low ASL activity, we speculate that δ1 expression may also predominate in the lens-fibrelike lentoid cells, further strengthening their molecular links with authentic lens fibre cells. If this is the case, then tissues expressing δ ectopically (e.g. fresh NR) may produce a higher proportion of δ2/ASL relative to δ 1 gene-products than is the case in lens fibres. This point requires further confirmation at the molecular level, using δ1-versus δ2-specific probes (Parker et al. 1988) and/or monoclonal antibodies specific for the δ1 and δ2 polypeptides.

ASL is one of the major urea cycle enzymes, and as such is known to be expressed in mammalian and adult amphibian liver tissues (both groups being ureotelic). Consistent with this, we find a δ-related protein of 48–50×10³/W_r in both mouse and Rana liver extracts, as well as in commercial bovine ASL (Fig. 4). Interestingly, we find a similar δ-related protein and significant ASL activity in Rana lens (Fig. 4, Table 1). We have not so far been able to immunoselect this protein (perhaps because Rana ASL differs significantly from chick δ in antigenicity), so we cannot as yet directly link these two observations. Nevertheless, the occurrence of ASL in Rana lens might help to explain earlier reports of δ-related polypeptides in the lenses of non-avian and non-reptilian vertebrates (e.g. de Pomerai et al. 1984). The metabolic function of δ2/ASL gene products in non-lens tissues (such as embryonic NR) remains to be elucidated.

 
• NR

  neural retina;

 
• ASL

  argininosuccinate lyase;

 
• SDS

  sodium dodecyl sulphate;

 
• PAGE

  polyacrylamide gel electrophoresis;

 
• UV

  ultraviolet light;

 
• VRCs

  vanadyl ribonucleoside complexes;

 
• BSA

  bovine serum albumin.

The authors would like to thank Dr J. Piatigorsky for supplying the pδCrl7 probe, Mr A. Brown and Mr R. Searcy for preparing the photographs, Mrs B. Williams for typing the manuscript, and the SERC for financial support (Grant No. GR/D45918 to Dr D. de Pomerai) and for a postgraduate studentship to M.McL.