Definition of 5' and 3' structural boundaries of the chromatin domain containing the ovalbumin multigene family.

Hen oviduct nuclei were subjected to pancreatic DNase I treatment under conditions known to preferentially degrade transcriptionally active genes (Weintraub, H., and Groudine, M. (1976) Science (Wash. D. C.) 93, 848-856). The ovalbumin gene, its structurally related genes, X and Y, and the spacer and flanking DNA were all found to exist in a DNase I-sensitive configuration. The DNase I-sensitive region was extended more than 20 kilobases beyond the 5' end of the X gene and approximately an equal distance beyond the 3' end of the ovalbumin gene before it became DNase I-resistant. The transition from a DNase I-sensitive to a -resistant conformation in oviduct chromatin occurred in a gradient fashion with 10 kilobases of DNA. Thus, ovalbumin and its related genes, X and Y, exist in a 100-kilobase DNase-sensitive domain in the oviduct tissue. In contrast, the entire domain was resistant to DNase I in spleen, liver, and erythrocyte nuclei. When the transcription of ovalbumin, X, and Y genes was eliminated by the withdrawal of hormone from estrogen-stimulated chicks, the entire domain remained in a DNase I-sensitive configuration. We conclude that DNase I-sensitive domains may provide the structural capability for gene expression and appear to be a result of the differentiation process since they are cell-specific and contain potentially expressible genes of that cell type. Repetitive sequences within this domain have been mapped and the possible relationship of these repetitive sequences to the DNase I-sensitive structure is discussed.

In contrast, the entire domain was resistant to DNase 1 in spleen, liver, and erythrocyte nuclei. When the transcription of ovalbumin, X, and Y genes was eliminated by the withdrawal of hormone from estrogenstimulated chicks, the entire domain remained in a DNase I-sensitive configuration. We conclude that DNase I-sensitive domains may provide the structural capability for gene expression and appear to be a result of the differentiation process since they are cell-specific and contain potentially expressible genes of that cell type. Repetitive sequences within this domain have been mapped and the possible relationship of these repetitive sequences to the DNase I-sensitive structure is discussed.
Numerous studies in recent years have f m l y established that expressible genes are packaged into chromatin differently as compared to regions of the DNA which are genetically repressed. In particular, genes which are transcriptionally active, or which have the potential for rapid expression in response to appropriate inducers, have been shown to exhibit a preferential susceptibility to cleavage by nucleases. Such genes include globin (1-3), ovalbumin (4)(5)(6)(7)(8), vitellogenin (9), insulin (lo), immunoglobulin (11,12), histone (13), protamine (14), ribosomal RNAs (15)(16)(17), and a variety of integrated genes for viral proteins (18)(19)(20). The enhanced sensitivity is generally manifested either as a rapid reduction of the entire * This work was supported by National Institutes of Health Grants HD-08188 and HD-07495-08. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ Present address, Clinic Chemistry Section, Mayo Clinic, Rochester, MN. sequence under study into small nonhybridizable fragments following extensive digestion in situ, or by specific cleavage at a few uniquely localized hypersensitive sites in the immediate vicinity of the gene under conditions of extremely mild digestion (21). In most of these studies, the sensitivity of specific genes appears to be correlated with the differentiated state of the cell rather than transcriptional activity per se since the acquisition of sensitivity has been seen to precede expression in some developmental programs (12,22) and to persist after cessation of expression in others (1,2). Moreover, inducible genes generally maintain their overall nuclease sensitivity as well as their characterisbic pattern of hypersensitive cleavage sites even in the uninduced state (8,23,24), even though subtle modulations in sensitivity attributable directly to the transcription process have been reported (5, 21, 25). Considering the variety of different genes and systems studied, and because no expressible gene has been reported to exist in a nuclease-resistant state, acquisition of a nuclease-sensitive conformation appears to be a general prerequisite to the potentiation of eucaryotic gene expression.
Using cloned DNA fragments as specific hybridization probes, our laboratory has been investigating the nuclease sensitivity in oviduct nuclei of defined regions of the chicken genome containing either the ovomucoid gene or the cluster of ovalbumin-reiated genes (comprised of the X, Y, and ovalbumin genes). These genes, whose DNA structures have been well characterized (26-29), are of particular interest because they are induced in tubular gland cells of the oviduct by a direct action of steroid hormones (28,(30)(31)(32)(33). Using the technique in which isolated nuclei are digested with DNase I to render 15-20% of the DNA acid-soluble (1,4), we have recently observed that the DNase I-sensitive conformation extends well beyond the boundaries of the transcription units into the sequences which flank the 5' and 3' ends of the genes in the chicken oviduct. In the case of the ovalbumin-related gene cluster, a 54-kb' length of DNA containing the three genes as well as the nontranscribed spacer and flanking sequences was found to exist in a DNase I-sensitive conformation (6). These, and similar observations by Stalder et al. (22) on chickenglobin genes support the concept that active genes, along with their flanking sequences, are organized into DNase I-sensitive domains.
The primary purpose of the experiments described in the present study was to obtain additional DNA probes further distal to the X and ovalbumin genes so that the precise dimensions of the active domain in oviduct chromatin could be determined. Having identified the boundaries of DNase I sensitivity, we wished to verify that selected regions of the domain exist in an insensitive conformation in some repre-

Multigene Families Exist in Chromatin Structural
Domains sentative tissues not expressing the oviduct-specific genes. We have also investigated the behavior of the transition regions following shutdown of gene transcription upon withdrawal from estrogen. Finally, we present a preliminary identification of repeated DNA sequences occurring throughout the domain and discuss the implications of such organization for the differentiative process.

EXPERIMENTAL PROCEDURES
Cloning of the DNA in Ovalbumin Domain-All the clones shown in Fig. 1 except CL734 were isolated from a chicken-DNA library (obtained from Drs. d. D. Engel, J. B. Dodgson, R. Axel, and T. Maniatis) prepared using Charon 4Ah DNA vector and chicken DNA that was partially digested by Hue 111 and Alu I and ligated to synthetic Eco RI linkers (34). Clone CL734 was obtained from a separate library constructed using Charon 4Ah DNA as vector and chicken DNA partially digested with Eco RI. Fragments are referred to in the text by their clone number followed in parentheses by their size in kilobases. Only those fragments referred to in the text are specifically designated by size on the map. The isolation and characterization of clones CL64, CL78, and CL36, as well as the localization of the X, Y, and ovalbumin genes, have been previously described (28). Clone CL125 was obtained by screening the library with a 32Pnick-translated 2.2-kb 5' antepenultimate fragment from CL36. Likewise, CL125 (2.3) was used as a probe to obtain CL252. From CL252, a 5"terminal 1.6-kb fragment (from CL252 (9.0) was generated by cleavage with Xho I and used as a probe to isolate CL167. Clone CU86 was obtained by screening with CL167 (8.5). In the 3' direction, CL57 was obtained by probing with the natural 9.2-kb fragment containing the 3' end of the ovalbumin gene (27), while CL734 was isolated by screening with the 0.6-kb 3' penultimate fragment of CL57. The clones indicated here are only a subset of those actually obtained since multiple clones with various degrees of overlap were generally isolated at each screening. Restriction mapping and confirmation of overlap by Southern analysis were according to established procedures (28).

DNase Z Sensitivity of Selected Sequences within the Ovalbumin
Domain in Hen-Oviduct Nuclei-Nuclei were isolated from the magnum portion of laying hen oviducts as previously described (6). Nuclei were suspended at a DNA concentration of approximately 1 m g / d in buffer containing 0.01 M Tris-HC1 (pH 7.5), 0.01 M NaC1, 0.003 M MgClr, and 0.35 M sucrose and digested at 37 "C with pancreatic DNase I (Worthington) until 15-20% of the DNA was rendered soluble in cold 7% perchloric acid. Purification of the DNA, including digestion with proteinase K, aIkaline hydrolysis at 68 "C, and gel filtration have been described in detail (6). Control DNA was extracted from hen-oviduct or hen-liver nuclei (comparable results were obtained with either preparation) not subjected to in situ nucleolysis and, following deproteinization, was subsequently sheared enzymatically to generate a DNA-fragment distribution comparable in size to the DNase I-digested preparation. Specific hybridization probes were prepared by restricting appropriate DNA clones ( Fig. 1) with Eco RI (or in the case of CL734, Eco RI + Xho I), isolation of individual restriction fragments by preparative electrophoresis in 1% agarose gels and absorption to glass bead after dissolution in 6 M NaC104, and elution from glass bead with 10 mM Tris-HCI, pH 7.5, and 1 mM EDTA after washing with 70% cold ethanol and then labeling with tritiated nucleotides (New England Nuclear) to high specific activity  or -control DNA), and 600-800 cpm of nick-translated DNA probe. Samples were denatured at 100 "C for 5 min and incubated at 68 "C for intervals ranging from 10 min to 24 h. Hybrid formation was determined by digestion with S1 nuclease (Miles Laboratories, Inc., Elkhart, IN, 4800 units/vial) for 2 h at 37 "C, precipitation with cold 10% trichloroacetic acid, and collection of trichloroaetic acid-precipitated material on nitrocellulose filters (Millipore Corp., Bedford, MA) which were dissolved in methylcellosolve and counted in Aquasol. Probe self-annealing was monitored in parallel vials containing 8 mg/ ml of yeast RNA, and all vials were corrected for this effect which rarely exceeded 15-20% of the S1-resistant counts/min. Equivalent Cot values have been plotted.
Detection of Repeated DNA Sequences in the Ovalbumin Domain-The nine clones shown in Fig. l were restricted with Eco RI (except CL734 which was restricted with Eco RI and Xho I), the restriction fragments resolved on a 1% agarose gel, and the fragments transferred to a nitrocellulose filter according to Southern (36). The fdter was hybridized for 16 h at 68 "C with 90 X lo6 cpm of denatured 32P-labeled total chick embryo DNA (nick-translated to a specific activity of 2 X 10' cpm/pg) in 2 X Denhardt's solution, 6 X SSC, 0.5% sodium dodecyl sulfate, 1 mM EDTA (pH 7.0), and 1 mg/ml of yeast RNA. The fdter was given six 30-min washes in 1 X SSC, 0.5% sodium dodecyl sulfate at 68 "C and exposed to Kodak XR-5 film with an intensifying screen.

RESULTS
DNase I Sensitivity of Sequences within the Ovalbumin Multigene Domuin-A map of the chicken ovalbumin gene domain showing the relative positions of various cloned fragments used in these studies is presented in Fig. 1. Screening procedures used to identify clones from chicken-DNA libraries, verification of overlap, and restriction mapping were performed using standard procedures (28). Individual restriction fragments were isolated on preparative agarose gels, labeled to high specific activity by nick-translation, and used as probes for Cd analysis in order to determine the concentration of their respective sequences in oviduct nuclei after nuclease digestion. Restriction fragments are identified in the text and figures by designating the clone number followed in parentheses by the fragment size in kilobases. Parallel hybridizations were conducted in which labeled probes were annealed in solution to driver DNA prepared from oviduct nuclei digested with DNase I to render 15-2076 of the DNA acid soluble (DNase I-digested DNA) and to control DNA isolated from nuclei not digested in situ but subsequently sheared to generate a population of DNA fragment comparable in size to that of the DNase I-digested preparation.
Our previous studies showed that the X, Y, and ovalbumin genes, along with their nontranscribed spacers and at least 8 kb of DNA flanking the 5' end of the X gene and 7 kb flanking the 3' end of the ovalbumin gene, are sensitive to DNase 1 (6) in hen-oviduct nuclei. The availability of more distal clones indicates that the sensitive conformation extends considerably further in either direction. The annealing data for fragment CL125 (6.0), whose 5' end flanks the X gene by approximately 17 kb, fragment CL78 (6.0), which includes the 3' half of the Y gene plus some flanking sequence (included here as a representative internal sequence), and fragments CL57 (1.3, 1.6, and 5.5), which collectively include an 8.4-kb segment which extends to a point 17   Left-hand panels, annealing curves. Right-hand panels, linear transformation of the companion annealing data. For ideal second order kinetics of reassociation, H / l -H i s proportional to Cot, where Hrepresents the fraction of probe existing as hybrid relative to the maximum hybridization observed. The slopes are proportional to the concentration of the labeled sequence in the respective driver DNA preparations so that the ratio of slopes provides an estimate of the degree of preferential nucleolytic sensitivity (4).

Fig. 2,
A-E, respectively. In order to estimate the concentration of the sequence of interest in the two driver DNA preparations, the annealing data have been replotted, as shown in the right-hand panels of Fig. 2, such that the concentration is proportional to the slope (4, 6). The concentration of each of these sequences is reduced approximately 2.5-to %fold in the DNase I-digested samples compared to the control preparations, a magnitude comparable to that reported by others using cDNA probes prepared against ovalbumin-messenger RNA (4, 8), and in accord with our previous studies assaying less distal flanking sequences as well as genomic sequences from within the transcription units of the three genes. By contrast, no preferential digestion of chicken /I-globin sequences is observed when a /I-globin-specific probe is included in the same type of analysis (Fig. 3F). These data thus confirm and extend our previous findings that flanking sequences well beyond the boundaries of the transcription unit can be packaged in a DNase I-sensitive conformation.
FIG. 3. Transition from a DNase I-sensitive to -resistant conformation at the 5' boundary of the ovalbumin domain in henoviduct nuclei. Restriction fragments CL486 (5.2), CL167 (8.5 and 2.6), and CL252 (9.0 and 2.6) (A-E, respectively) were labeled by nick-translation and hybridized either to DNase I-digested DNA prepared from hen-oviduct nuclei (0) or to control DNA from henliver nuclei (0). F, a labeled 6.1-kb restriction fragment containing chicken P-globin sequences (58) (not expressed in the oviduct) was annealed to these same driver DNA preparations. Hybridizations were performed as described in the legend to Fig. 2. Linearized annealing data have been plotted.

The 5' and 3' Boundaries of the Ovalbumin Domain-
Three more overlapping clones (CL252, CL167, and CL486) were obtained which extend an additional 35 kb beyond the 5' end of the CL125 insert. Isolated restriction fragments from these clones, ordered as indicated in Fig. 1, were labeled by nick-translation and used as probes for DNase I sensitivity of their respective sequences in hen-oviduct nuclei as described above. The hybridization data for five of these fragments are presented in Fig. 3. Fragment CL252 (2.6) (Fig. 3E), whose 5' end is located 19 kb 5' distal to the X gene, exhibits a sensitivity to DNase I comparable to that observed for the other fragments within the domain. (In a separate experiment, identical results were obtained using fragment CL125 (2.3) which overlaps CL252 (2.6) (data not shown)). However, fragment CL252 (9.0) (Fig. 3 0 ) shows a graded loss of sensitivity which becomes more pronounced for fragment CL167 (2.6) (Fig. 3C). Fragments CL167 (8.5) and CL486 (5.2) (Fig. 3, A  and B ) , which lie even further upstream, show only marginal sensitivity to DNase I and exhibit a behavior similar to the repressed P-globin sequence which was included in the same experiment and is shown here for comparison (Fig. 3F). In addition, fragments CL167 (1.4 and 4.1) and CL486 (3.6 and 6.1) were also assayed for sensitivity and the data were consistent with those obtained for the fragments presented in Fig.  3 (data not shown).
At the 3' end of the domain, clone CL734, which contains a chicken-DNA insert whose 3' end extends to a n Eco RI site approximately 27.5 kb downstream from the ovalbumin gene, was isolated. This insert contains an 11-kb Eco RI fragment which was subsequently trisected by Xho I into fragments of 3.6,2.3, and 5.1 kb in length. These fragments were ordered as indicated in Fig. 1 and used as probes for the nuclease-sensitivity assay. Fragment CL734 (3.6), whose 3' end is located 20 kb 3' distal to the ovalbumin gene, is fully sensitive to DNase I (Fig. 4.4). However, a graded decrease in sensitivity, analogous to that occumng at the 5' end of the domain, is observed for the more distal fragments, CL 734 (2.3 and 5.1) (Figure 4,   B and C). Thus, the total extent of DNase I-sensitive DNA comprising the ovalbumin domain in oviduct nuclei is slightly greater than 100 kb.

DNase I Resistance of the Ovalbumin Domain in Other
Tissues-We have previously shown that several sequences from the ovomucoid domain, as well as fragment CL64 (5.5) from within the ovalbumin domain (Fig. l), are resistant to DNase I in spleen, liver, and erythrocyte nuclei (6). Since the detection of discrete boundaries which flank the ovalbumin domain suggests the possibility that the alteration in chromatin structure occurring at these loci may function in some capacity to maintain and/or establish the active conformation requisite to gene expression, we wished to verify that these structural transitions are in fact unique to oviduct chromatin.
T o this end, spleen, liver, and erythrocyte nuclei were digested with DNase I. The DNase I-digested DNAs, along with a control preparation, were annealed to probes at the 5' (CL252 (2.6 and 4.1); CL167 (8.5)) and 3' (CL734 (5.1)) boundaries. As seen from the hybridization data in Fig. 5, the concentration of these sequences in the DNase I-digested and control preparations were indistinguishable, indicating the existence of a nuclease-resistant conformation for these regions in nuclei not destined to express the ovalbumin-gene family. Maintenance of the Ovalbumin Domain Following Estrogen Withdrawal-The globin gene retains its sensitivity to DNase 1 in transcriptionally inactive erythrocytes (1) as does the ovalbumin gene following withdrawal of steroid hormones (8). We have investigated the DNase I sensitivity of some selected sequences throughout the ovalbumin domain following hormone withdrawal, with particular emphasis on the 5' and 3' boundaries in order to determine if positional shifts in these outlying regions might occur during different states of modulation by steroid hormones. Three-week-old immature chicks were injected daily for 2 weeks with diethylstilbestrol and subsequently withdrawn from hormone for 3.5 days, an interval shown to be sufficient for complete inactivation of the ovalbumin gene in an in vitro nuclear transcription/fiiter hybridization assay (37). Nuclei were isolated from chick oviducts before and after withdrawal and transcribed in vitro to verify that withdrawal was complete (data not shown). Nuclei isolated from withdrawn oviducts were digested with DNase I under standard conditions. The DNase I-digested DNA was hybridized in parallel with a control preparation against probes from the 5' boundary (CL167 (8.5 and 4.1) (Fig.  6, A -C ) ) , a probe from the central region of the domain (CL78 (6.0) (Fig. 6 D ) ) , and probes from the 3' boundary (CL734 (3.6 and 5.1) (Fig. 6, E and F ) ) . These data are comparable to those obtained using oviducts from laying hens and indicate a functionally rigid domain having dimensions that are invariant and independent of the transcription level of the included genes. The magnitude of discrimination for sequences in the sensitive portion of the domain is slightly less in oviduct nuclei from withdrawn chicks compared to laying hens (Fig. 8), a result likely to be attributable to a lower percentage of tubular gland cells in the former (8) and probably does not represent a fundamental structural difference in the two domains.
Identification of Repeated Sequences in the Ovalbumin Domain-An obvious question concerns what might be involved in establishing and maintaining such a domain. Since DNA-sequence recognition might ultimately be involved, and if the organization of expressible genes into domains is a common phenomenon resulting in the existence of hundreds or thousands of active domains within a given nucleus, repetitive DNA sequences might be expected to play some role. We have, therefore, screened the entire domain for the presence of repetitive sequences to determine if any pattern emerges in relation to the transition regions. The DNA clones indicated in Fig. 1 were restricted with Eco RI (except for CL734 which was restricted with Eco RI and Xho I), electrophoresed on agarose gels, and transferred to nitrocellulose filters (36). The filters were hybridized against nick-translated total chick-embryo DNA under conditions in which unique sequences fail to give a detectable signal (38). The agarose gel stained with ethidium bromide along with its companion autoradiogram are displayed in Fig. 7, A and B, respectively. The restriction fragments fall into three categories. The first yields no detectable signals even after prolonged exposure. A second class yields lower intensity signals (CL167 (2.6, 4.1); CL252 (9.0, 2.6, 6.0); CL125 (2.3, 6.0); CL36 (1.7, 4.6); and CL78 (7.2), the latter three fragments containing repeated sequences previously identified) (39). A third class is characterized by significantly more intense signals (CL125 (4.2), CL36 (3.5), and CL57 (5.5)). Since fragment CL36 (3.5) is  Fig. 1 were restricted with Eco RI (except CL734 which was restricted with Eco RI and Xho I), the restriction fragments resolved on a 1% agarose gel, and the fragments transferred to a nitrocellulose filter and hybridized to nicktranslated total chick-embryo DNA. A, the agarose gel stained with ethidium bromide. B, autoradiogram after transfer to nitrocellulose and hybridization against labeled chick genomic DNA.
Restricted clones are identified in the figure. Higher molecular weight phage fragments common to alI clones are not shown. Clone CL734 was run on a separate gel from that of the others displayed here but because common clones were included on both gels, the exposure has been adjusted so that the relative intensity of CL734 (5.1) is directly comparable to the other fragments. The larger labeled fragments in CL36 not corresponding to any major insert bands are partial digestion products containing the intensely hybridized CL36 (3.5) fragment.

I '
FIG. 8. Summary of data of the ovalbumin domain. The relative sensitivity of a given sequence is defined as the ratio obtained when the slope of the transformed annealing curve generated with control DNA as driver is divided by the slope when DNase I is driver.
Each data point represents an independent experiment using a specific restriction fragment as probe and is positioned on the map at the midpoint of that fragment. The relative sensitivity has been assayed in nuclei prepared from hen oviduct (O), estrogen withdrawn chick oviduct (0). hen liver (A), chick spleen (A), and hen erythrocytes (0). The midpoints of restriction fragments containing repeated DNA sequences (Fig. 7) are indicated by arrouq the larger of which represent the more highly reiterated sequences.
contained entirely within CL125 (4.2), their respective signals most likely score for the same repeated sequences.
With regard to these two more highly reiterated sequences, to date we have only identified the restriction fragments which contain them and have not yet characterized them with respect to precise location, size, possible relatedness, and actual reiteration frequency in the genome. Although the function of these sequences is not known, their proximity to the 5' and 3' transition regions is striking. It is tempting to speculate that they could function in some capacity to define the limits of the active domain. In this respect, it is of interest that one of the rat-preproinsulin genes is included in a 12-kb length of DNA flanked by 1.1-kb inverted repeats (40) while approximately 35 kb of rabbit DNA containing a cluster of four pglobin genes is flanked by 1.4-kb inverted repeats (39).
The Oualbumin-Gene Domain as a Functional Unit of Genetic Expression-The nuclease-sensitivity data presented above are summarized in Fig. 8. Also included in this figure are points derived from previously reported data (6). Due to the limited resolving space of the assay in terms of modest Cot shift even in the region of greatest sensitivity and because the genome is being probed in discrete "quanta" of finite length, the precise loci on the 5' and 3' flanking DNA a t which the onset of a DNase I-resistant conformation commences can only be approximated. However, it is evident that the X, Y, and ovalbumin genes are situated in a DNase I-sensitive domain having a size of 100 kb with approximately 20 kb of DNase I-sensitive DNA flanking both the 5' end of the X gene and the 3' end of the ovalbumin gene; the transitional resistant region covers about 10 kb a t both the 5' and 3' ends of the domain. If the nuclease-sensitive DNA is organized into nucleosomes with a 200-base pair periodicity (41,42), the ovalbumin domain would be packaged as approximately 400 nucleosomes in uiuo.

DISCUSSION
Our results demonstrate that in chick oviduct cells the ovalbumin, X and Y genes, and the spacers between these genes, as well as the flanking sequences, all exist in a DNase I-sensitive structural domain. This domain is surrounded by a DNase I-resistant chromatin configuration. The transition from a sensitive to a resistant structure occurs in a graded fashion resulting in a moderately broad (-10-kb) boundary. Such breadth may accurately reflect the structure existing in

Multigene Families
Exist in Chromatin Structural Domains each tubular gland cell, but the possibility cannot be eliminated that the transitions are actually much sharper in individual nuclei but that the precise location tends to vary within limits from nucleus to nucleus, In any event, the transition appears to generate a highly "inactive" conformation since the most 5' distal sequences exhibit a sensitivity to DNase I similar to that observed for the P-globin gene which is not expressed in the oviduct and presumably exists in a fully repressed state.
There is a related alternative that also merits consideration. Structures similar in size to the 100-kb ovalbumin gene domain have been previously identified in the nucleus. These include the chromomeres of polytene chromosome (average DNA size of 100 kb) (43), as well as the DNA loops observed after deproteinization in metaphase chromosomes (30-90 kb) (44), and in interphase chromatin (85 kb average) (45). Moreover, these loops appear to be anchored to a nuclear scaffold or matrix (46, 47) in an apparently nonrandom manner with respect to specific DNA sequences (48, 49). Conceivably, the ovalbumin domain may represent such a structure with the matrix-interaction sites possibly occurring in the vicinity of the transition or nuclease-resistant regions. The possible role of the highly reiterated sequences which flank the gene cluster in defining the limits of the ovalbumin domain as discussed above might somehow key appropriate interactions with components of the matrix. Preliminary results consistent with this viewpoint indicate the presence of an adjacent DNase I-sensitive domain further upstream to the 5' side of the ovalbumin domain, such that the 5' resistant region may in fact resemble a node separating consecutive domains.' In the simplest case, the DNA not associated directly with the matrix could be packaged either into nuclease-sensitive (extended or eucromatic?) or -insensitive (condensed or heterochromatic?) loops. Nuclease resistance could then be conferred to a given sequence either because of its proximity to the matrix (5' and 3' regions flanking the ovalbumin domain) or its inclusion in a nuclease-resistant loop (the repressed P-globin gene). A possible analogy between the globin domain and lampbrush chromosomes has also been discussed (3).
Under our conditions of extensive digestion, the X, Y, and ovalbumin genes appear equally sensitive to DNase I even though they are transcribed at significantly different rates (6, 28). This finding agrees with that of Garel et al. (41), who originally demonstrated, using a heterogeneous probe, that even genes transcribed at very low rates are sensitive to DNase I. Considering the multiplicity of sequences simultaneously assayed in this study along with the increasing number of specific genes analyzed by others, it would appear that the organization of expressible genes into DNase I-sensitive domains of discrete size may be an absolute prerequisite to the expression of any eucaryotic gene.
It also appears that the ovalbumin domain functions as a discrete unit of genetic expression in the differentiated tubular gland cell. Although the three genes are transcribed at different rates, all are coordinately induced in the chick oviduct (28). They can also be induced by other nonestrogenic steroids but the gradient of differential expression is maintained (50). These observations suggest that the ovalbumin domain as a whole can undergo cycles of activation and inactivation, respectively, in response to hormonal induction or withdrawal, but that the relative level of expression of the individual genes in the activated state is more or less invariant. That domains in general may function as discrete units of genetic expression is suggested by the fact that, although structurally related genes can be either tightly clustered or widely dispersed throughout the genome, the clusters described to date invar-* G. Lawson, unpublished observation.
iably contain genes serving a common function (51, 52). The level of individual gene expression is, however, apparently modulated by additional factors other than just their sequestration into active domains since unique genes can be expressed at very different levels in different tissues (2), and in the case of the chicken P-globin family, the temporal expression of the embryonic and adult genes is dissociated even though they are probably contained within the same domain This definition of the domain is an operational one dependent upon the conditions used to probe chromosome structure. We have observed no detectable difference in the level of DNase I sensitivity exhibited by transcribed segments of the X, Y, and ovalbumin genes and their nontranscribed flanking sequences following extensive fragmentation of the DNA (ie. 15-20% rendered perchloric acid-soluble) ( 6 ) . However, subtle differences in the nuclease-sensitive state have been observed by others employing less extensive digestion. Applying a limited nicking and filter blotting assay (25) in which the level of digestion was several orders of magnitude less than the one utilized here, Stalder et al. (22) have been able to detect a subdomain which comprised only the transcribed gene and its immediate surrounding regions within the larger globin domain similar to that reported here. The transcribed genes, which exhibit the greater sensitivity, appear to be associated with nucleosomes containing hypomethylated DNA and high mobility group 14 and 17 (HMG 14 and 17) nonhistone proteins (53, 54). Differing sensitivities have also been observed for the ovalbumin gene in the hormonally stimulated and withdrawn states (5) between the actively transcribed ovalbumin gene in the oviduct and the once-expressed &globin gene in the erythrocyte (7). Additionally, in cooperation with John Anderson and his co-workers, we have observed differential sensitivity to micrococcus nuclease between the highly transcribed ovalbumin gene and the more modestly transcribed X and Y genes.3 Finally, alterations in chromatin structure between transcription units and interspersed nontranscribed spacers have been studied by nuclease digestion of ribosomal genes (16), electron microscopy studies of ribosomal genes (55, 56), and Balbioni rings (57).
We relate our findings to those above by assuming that the level of nucleolysis employed here, which generates singlestranded DNA fragments ranging in size from 10 to no greater than a few hundred nucleotides (6), results in rapid obliteration of the more labile substructures which consequently go undetected. We thus postulate that the domain defined here may represent a primary level of structural organization upon which the other subdomains are constructed.
Our results are consistent with the model presented in Fig.  9. In this model, we propose that during the differentiation process, differentiation-specific genes in various cell types are incorporated into loops or DNase-sensitive domains. In the absence of additional modulators of transcription, these looped out domains remain unexpressed. Upon interaction with inducers such as hormones for the ovalbumin-gene family in chick oviduct, the genes within the domain become transcriptionally active and remain DNase I-sensitive. Within this large domain, transcribed regions of DNA may be segregated into subdomains which are even more sensitive to DNase. The degree of nuclease sensitivity among these subdomains may, in fact, be related to the level of transcription. Upon the removal of inducers, the transcriptional activity for these genes disappears although DNase sensitivity within this domain remains. Thus, a critical step in the process of cellular differentiation may involve a sequestering of the appropriate