Nuclear matrix interactions within the sperm genome.

Analysis of the haploid-expressed human PRM1 PRM2 TNP2 genic domain has revealed two regions of attachment to the sperm nuclear matrix. These sperm nuclear matrix attachment regions delimit the DNase I-sensitive domain of this haploid-expressed locus. The domain is intermediately associated with but not attached to the nuclear matrix. DNase I-sensitive genes within the mature sperm nucleus, such as protamine 1, protamine 2, transition protein 2, α-globin, and β-actin, display this intermediate affinity for the sperm nuclear matrix. This may denote their role in templating the male genome prior to fertilization, thus ensuring the formation of a viable male pronucleus during early embryonic development.

For many years the nuclear matrix received little attention, as it was thought to act merely as a structural element (1). It has now been suggested that the nuclear matrix may play a key role in genome organization and gene potentiation (2). As in the somatic nucleus, chromatin within the male gamete is organized into discrete loops, bound at the base by regions of attachment to the nuclear matrix (3). These loops differ from their somatic counterparts with respect to the packaging of their DNA (4) and their average size. Loops within the sperm nucleus are ϳ27 kb 1 in size (5) compared with ϳ60 kb in all other types of cells studied to date (6). We have termed these sperm nuclear matrix attachment regions (SMARs) (7). The somatic nuclear matrix has come under intense study, as actively transcribed genes have been shown to be associated with the nuclear matrix (8). Somatic nuclear matrix attachment regions (MARs) have been identified in or near introns (9), enhancers (10), origins of replication (11), and sites of transcription initiation (12), as well as other regulatory elements (9). MARs have also been identified at the ends of the DNase I-sensitive domain in numerous loci (13,14) and shown to facilitate position-independent gene activity (15). The function of the sperm nuclear matrix is comparatively unknown.
An ϳ40-kb region of human chromosome 16p13.13 has recently been sequenced in its entirety and shown to contain the genes for the sperm-specific protamine 1, protamine 2, and transition protein 2 proteins (16). DNase I sensitivity analysis has delineated the boundaries of the domain in the mature spermatozoan, and transgenic analysis has shown that this region of the genome contains all the elements necessary for the appropriate spatial and temporal expression of the genes of this cluster in a position-independent, copy number-dependent manner (17). To characterize structural elements that mediate this response, we have identified regions of genomic interaction with the sperm nuclear matrix. Further, we demonstrate that a specific subset of both haploid-specific and constitutively expressed genes are associated with the mature sperm nuclear matrix. These genes assume an altered structural conformation as evidenced by their increased sensitivity to DNase I. Thus, the mature sperm genome is organized in a specific non-random manner. This could provide the means to template the male genome for ordered protamine replacement immediately subsequent to fertilization.

MATERIALS AND METHODS
Physical characterization of each of the candidate MARs employed nuclei prepared from frozen sperm essentially as described (19). Nuclei were resuspended in 50 mM HEPES, pH 7.5, buffer containing 10 mM NaCl, 5 mM MgOAc, and 25% glycerol, at ϳ1 ϫ 10 7 /ml, and then used immediately or stored flash frozen at Ϫ80°C. DNA halos were prepared from fresh or frozen sperm nuclei as described (5). In brief, sperm nuclei were mixed with an equal volume of 2 M NaCl buffered with 25 mM Tris, pH 7.4, and then pelleted at 4°C for 30 min at 1,600 ϫ g. The pellet was resuspended in 200 l of 25 mM Tris, pH 7.4, buffer containing 2 M NaCl and then adjusted to contain 10 mM dithiothreitol. The nuclei were then incubated on ice for 30 min. The resulting halos were centrifuged at 4°C for 30 min at 1,600 ϫ g and then resuspended in 50 mM Tris-HCl, pH 7.5, buffer containing 100 mM NaCl and 10 mM MgCl 2 . Aliquots were stained with propidium iodide and then visualized by fluorescent illumination using a Leitz DIAPLAN microscope. The remaining halo DNA was subsequently digested with BstXI, EcoRI, HindIII, or StyI for 4 h at 37°C. Successful restriction enzyme digestion was assayed by the inability to amplify across known sites. Following digestion, an equal aliquot of 4 M NaCl was added, and the samples were incubated for an additional 10 min at 37°C. The loop and matrix fractions were then separated by centrifugation for 30 min at 9,000 ϫ g at 4°C. The fractions thus separated were subsequently purified using Prep-A-Gene matrix (Bio-Rad) and then resuspended in deionized water. PCR amplification was performed on both the loop and matrix-associated fractions utilizing primer pairs directed to the PRM1 3 PRM2 3 TNP2 locus, many of which have been described previously. 2 PCR was maintained within the linear range of amplification. DNA halos were prepared from HeLa cells essentially as described (21), digested to completion with HindIII, and then treated as described above for sperm halos.

RESULTS AND DISCUSSION
To begin to elucidate the elements necessary to potentiate this domain, candidate regions of sperm nuclear matrix association within the PRM1 3 PRM2 3 TNP2 biological locus were identified utilizing a computational strategy. Characteristic MAR motifs were gathered from the literature (7,22) and then expressed as unique sequence patterns as described (18). In this manner, the ϳ40-kb sequence containing the PRM1 3 PRM2 3 TNP2 biological locus was queried for the presence of * This work was supported by Grant 1R01HD2850401A1 (to S. A. K.) from the National Institute of Child Health and Development and Grant EDUD-US93015 from SUN microsystems. 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.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EMBL Data Bank with accession number(s) U15422.
‡ Supported in part by the Dean's postdoctoral recruitment fellowship. 1 The abbreviations used are: kb, kilobase(s); SMAR, sperm nuclear matrix attachment region; MAR, somatic nuclear matrix attachment region; PCR, polymerase chain reaction; bp, base pair(s). 2 Primer sequences, PCR conditions, and the PRM1 3 PRM2 3 TNP2 domain sequence will be made available at the internet address "http://compbio.med.wayne.edu/". various sequence patterns indicative of MARs. Motifs were then weighted according to their expected frequency in a random sequence of the same base composition as that of the sequence queried. A weighted sum was subsequently applied to each region along the locus using a sliding window of 1000 bp with a 100-bp step size. The results are presented graphically in Fig. 1. Regions above a likelihood of 50% were considered as candidates to have strong nuclear matrix binding potential. This computer analysis predicted two SMARs centered at nucleotide positions 8,175 and 34,100 (Fig. 1). These potential SMARs were similar to those previously identified in this locus (7) and were used to guide their physical identification. DNA "halos" were prepared by extracting sperm nuclei with a high ionic strength reducing buffer (5). This displaced the histones and protamines from the chromatin, while leaving the DNA attached at discrete points to the intact nuclear matrix. The resulting halo structures were then stained with propidium iodide and visualized by fluorescence microscopy as shown in Fig. 2. The intact nuclei stained in a uniform manner, consistent with tightly packaged sperm chromatin, while the halo structures showed a more dispersed pattern of staining. Regions of the sperm chromatin that remained associated with the nuclear matrix possessed a brightly staining center, while the unassociated loop DNA stained dimly. This was manifested as a broad fibrous "halo" of fluorescence surrounding the brightly stained nuclear matrix.
To separate the nuclear matrix-bound and unbound DNA, halos were digested with various restriction endonucleases, and then the nuclear matrix-bound DNA was pelleted. Both fractions were purified and then subjected to PCR amplification using unique sets of primers targeted to discrete regions of the haploid-expressed PRM1 3 PRM2 3 TNP2 locus and the somatic expressed ␤-globin locus (Fig. 3). The distribution of each amplicon showed one of three patterns. The majority, i.e. at least 80% of the non-matrix-associated loop DNA, partitioned to the supernatant. Similarly, greater than 80% of the matrix-attached DNA partitioned with the nuclear pellet. In contrast, intermediately matrix-associated DNA partitioned into both the supernatant and pellet (30 -70%). Regions that lay outside the DNase I-sensitive domain localized consistently to the non-matrix-associated loop fraction, as did the DNase I-insensitive ␤-globin locus. Regions within the domain were intermediately associated with the nuclear matrix. This intermediate affinity for the sperm nuclear matrix is similar to that observed for the human ␤-interferon locus (23). The region surrounding and including the potentiated ␤-interferon gene shows weak nuclear matrix association and is bounded by points of stronger contact with the nuclear matrix. Regions near the ends of the PRM1 3 PRM2 3 TNP2 DNase I-sensitive domain were bound to the sperm nuclear matrix. These appear to be localized in a manner similar to the MARs of the chicken lysozyme (13) and human apolipoprotein B (14) loci. The 5Ј region of attachment to the sperm nuclear matrix was bounded by positions 8,818 -9,760, while the corresponding 3Ј region was bounded by positions 32,586 -33,536. The strong attachment to the nuclear matrix of these ϳ950-bp regions at the ends of the PRM1 3 PRM2 3 TNP2 DNase I-sensitive domain suggests the presence of a sequence-dependent MARlike element. However, these regions do not share extensive similarity.
Regions of intermediate nuclear matrix association are likely  (20). DNA halos were prepared, digested with various restriction endonucleases, separated into their loop (supernatant) and nuclear matrix-bound (pellet) fractions, and purified. PCR primer pairs that span specific regions of the PRM1 3 PRM2 3 TNP2 domain are shown below the ruler as small black boxes. 2 Primer pairs delimit the corresponding amplicons within the loop and nuclear matrix fractions. A PCR primer set directed to the ␤-globin locus was used as a non-matrix-associated control. This same region of the ␤-globin locus contains a somatic MAR, as shown in HeLa nuclei. Nuclear matrixbound restriction fragments in which greater than 80% of the amplicon partitioned with the matrix-bound fraction are identified as black boxes. Nuclear matrix-associated fragments are indicated by gray boxes for those amplicons that partitioned (30 -70%) within both fractions. Nonmatrix-associated fragments are demarcated by open boxes for those amplicons that comprised from 0 to 20% of the matrix fraction. Large restriction fragments that contain the SMARs often showed sterically reduced localization to the matrix fraction. Sites of attachment to the nuclear matrix are denoted as stars. Matrix association for the StyIdigested sample could not be ascertained for the ␤-globin locus, as there is a StyI site between the ␤-globin primers.
to reflect local differences in the organization of sperm chromatin. It has been shown that approximately 15% of the chromatin in human sperm remains histone-bound rather than undergoing protamine replacement (24). The intermediate association that is observed (Fig. 3) may reflect differential affinity of the sperm nuclear matrix for histone-bound chromatin as compared with protamine-bound chromatin. This is consistent with previous DNase I-sensitivity data of the PRM1 3 PRM2 3 TNP2 domain in human sperm (17), reflecting increased accessibility of this segment of the genome to the exogenous nuclease. Accordingly, DNase I sensitivity may be correlated with the degree of interaction of each gene with the sperm nuclear matrix. To test this hypothesis, mature spermatozoa loop and matrix-bound DNAs were subjected to PCR analysis using primer sets directed toward numerous well characterized loci throughout the human genome. As shown in Fig. 4, PCR analysis of the regions containing the PRM1, PRM2, and TNP2 genes showed that these genes were associated with the nuclear matrix. Similarly, amplification of regions of the ␣-globin HBA2 and ␤-actin genes, which are also DNase I-sensitive in mature spermatozoa, 3 also showed an association with the sperm nuclear matrix. In contrast, the ␤-globin, acrosin, and PGK-1 and PGK-2 genes, all of which are DNase I-insensitive in terminally differentiated mature spermatozoa, 3 showed no association with the sperm nuclear matrix. In somatic nuclei, DNase I sensitivity has been shown to correlate directly with the potentiation of genes for transcription. While there does not appear to be any transcription in mature sperm, the intermediate association of the DNase I-sensitive regions with the sperm nuclear matrix may represent a means by which the paternal genome is imprinted for activation and/or templated for postfertilization protamine replacement. Both processes are necessary for the formation of a viable male pronucleus.
It is clear that MARs and SMARs share only limited sequence and organizational characteristics. For example, the mouse ␤-globin locus has been shown to contain a MAR that functions independently of the type of somatic cell (29). It always anchors that region of the genome to the somatic nuclear matrix. As shown for HeLa nuclei in Fig. 3, this property is shared with the human ␤-globin locus. However, unlike the organization within the somatic nucleus, this region clearly does not interact with the mature haploid sperm nuclear matrix. This is in corollary with that observed for the haploid sperm-specific PRM1 3 PRM2 3 TNP2 locus.
In accord with the data presented above and that of others, there must be more than one type of association with the nuclear matrix. It is reasonable to assume that there are at least four classes of nuclear matrix association, i.e. regulatory element-associated MARs, somatic boundary elements, haploid boundary elements, and structurally associated elements. Class 1 regulatory element-associated MARs possess an innate ability to be bound by the nuclear matrix as they can be identified by an in vitro competition assay (26). These MARs are not typically situated at the ends of the DNase I-sensitive domain. They have been localized to regions containing enhancers (10), origins of replication (11), and other regulatory elements (9) and may also represent regions where transcriptionally generated supercoiling is relaxed (2). Class 1 MARs likely contain specific consensus sequences recognized by cell-specific nuclear matrix proteins. In fact, the nuclear matrix protein NMP-1, which binds to specific sequences within the histone H4 gene, has recently been shown to be the transcription factor YY1 (27). However, most of these MARs probably do not act as promoters or enhancers themselves. Instead, proximity of the regulatory element to the matrix-associated region and the nuclear matrix may concentrate all of the diverse elements necessary for transcription. In light of the locus-specific regulatory sequence motifs and the array of cell-specific proteins within the nuclear matrix (28), it is possible that no single consensus sequence for the class 1 MAR will be identified.
Class 2 somatic boundary element MARs are localized to the ends of DNase I-sensitive domains and act as boundary elements in somatic nuclei. They may shield loci against inappropriate potentiation and silencing in multiple types of cells. The MARs of the chicken lysozyme locus that delimit the DNase I-sensitive domain have been shown to mediate position-independent expression (15). It has been suggested that end region MARs may regulate transcription by inducing negative superhelical torsional stress across the domains that they limit (2). A universal consensus sequence for this second class of MAR should become clear as more are identified and sequenced, since many loci possess cell type-independent end region MARs. The AT-rich MAR may be representative of this class.
The regions of matrix association described above for the haploid-specific PRM1 3 PRM2 3 TNP2 domain are representative of class 3 haploid boundary element nuclear matrix attachment regions, i.e. SMARs. This report is the first identification of a haploid-specific MAR. Like the class 2 MARs, SMARs seem to act as boundary elements, attaching the ends of chromatin domains to the sperm nuclear matrix. The validation of the computational model suggests that SMAR sequences resemble those of the class 2 somatic boundary element MARs. However, MARs and SMARs are not identical. Unlike class 3 SMARs, class 2 somatic MARs have been shown to be cell type-independent. For example, MARs of three developmentally regulated Drosophila melanogaster genes have been shown to exhibit identical binding profiles regardless of tissue type or developmental stage (10). Further, MARs from the ␤-globin locus remain constant throughout the induction of terminal differentiation of the erythroid progenitors (29), while MARs of the chicken histone genes have been shown to be retained throughout the cell cycle (30). These differences among the class 3 SMARs and the class 1 and class 2 MARs are highlighted in Figs. 3 and 4. Genes like ␤-globin that contain a somatic cell type-independent MAR (29) do not partition with the sperm nuclear matrix, and SMARs of the haploid-expressed PRM1 3 PRM2 3 TNP2 domain do not attach to the HeLa nuclear matrix (Fig. 3). As with the class 2 MARs, identification of a consensus sequence for SMARs will depend upon the identification and sequencing of multiple SMARs. The intermediate association with the sperm nuclear matrix of those genes that exhibit DNase I sensitivity in mature spermatozoa can be considered to exemplify a class 4 structurally mediated nuclear matrix association. This intermediate affinity for the sperm nuclear matrix may be similar to that observed for the somatic expressed ␤-interferon gene. However, in haploid cells, this cannot be identified using an in vitro competition assay, and it appears that it is not dependent on the presence of a consensus sequence. It is not known if this type of association reflects a structural parameter specific to sperm chromatin.
While the function of MARs has been discussed extensively, the biological role for nuclear matrix attachment and nuclear matrix association within the male haploid genome remains to be clarified. The class 3 end region SMARs, like the class 2 end region MARs discussed above, appear to act as boundary elements. It is not clear whether they shield from position effects, as has been shown for some class 2 MARs (15). Three independent lines of transgenic animals containing SMARs from the PRM1 3 PRM2 3 TNP2 locus have been shown to yield copy number-dependent, site of integration-independent expression (17). However, such expression can also be achieved by a locus control region, as exemplified by the ␤-globin locus (25). Whether the human PRM1 3 PRM2 3 TNP2 locus contains an locus control region and/or utilizes the SMARs as a means of locus control remains uncertain.
The haploid-specific SMARs and the intermediately associated regions described above represent two of at least four classes of nuclear matrix-associated regions. Further clarification of the classes and functions of various nuclear matrixassociated regions will prove both interesting and enlightening toward the study of the mechanisms of gene potentiation and paternal genome templating.