Reductions in linker histone levels are tolerated in developing spermatocytes but cause changes in specific gene expression.

H1 linker histones are involved in packaging chromatin into 30-nm fibers and higher order structures. Most eukaryotic cells contain nearly one H1 molecule for each nucleosome core particle. Male germ cells in mammals contain large amounts of a germ cell-specific linker histone, HIST1HT, herein denoted H1t, which is particularly abundant in pachytene spermatocytes. Despite its abundance in male germ cells and significant divergence in primary sequence from other H1 subtypes, inactivation of the H1t gene in mice showed that it is not required for spermatogenesis. Analysis of germ cell chromatin from H1t null mice showed that other H1 subtypes, especially the testis-enriched HIST1H1A, herein denoted as the H1a subtype, were able to compensate for the absence of H1t to maintain a normal total H1 to nucleosome core ratio. To disrupt the compensation, we generated H1t and H1a double null mice by two sequential gene-targeting steps in embryonic stem cells. Elimination of both H1t and H1a led to a 25% decrease in the ratio of H1 to nucleosome cores in double null germ cells. Surprisingly, the reduction in H1 did not perturb spermatogenesis or produce detectable defects in meiotic processes. Microarray analysis of gene expression showed that the reduced linker histone levels did not affect global gene expression, but it did cause changes in expression of specific genes. Our results indicate that a partial reduction in linker histone-nucleosome core particle stoichiometry is tolerated in developing male germ cells.

Histones are small, basic proteins that are unique to eukaryotes and found in all eukaryotic nuclei. Histones are involved in packaging DNA into a compact structure called chromatin. There are five major classes of histones, the core histones H2A, H2B, H3, and H4, and the linker histone H1. Two molecules of each of the core histones constitute the pro-tein octamer of the nucleosome core particle. H1 histones bind to the linker DNA between nucleosome core particles and facilitate the folding of nucleosomes into the 30-nm chromatin fiber and higher order chromatin structures (1,2). Therefore, they would be expected to have profound effects on many cellular processes that require access to the genome.
A general feature of histone proteins is that their amino acid sequences are conserved. However, the H1 family of histones is the most diverse class, with multiple subtypes in most organisms (3). In mice eight H1 subtypes have been described (4,5), including the "somatic" subtypes H1a through H1e, the replacement subtype H1FO (here H1 o ), the oocyte-specific H1FOO, here denoted as H1oo, and the testis-specific H1t. These subtypes are also present in humans, and the genomic organization of the genes encoding the H1 subtypes appears to be very similar in humans and mice (6,7).
Although all of the mouse H1 subtypes have the basic tripartite structure of metazoan H1s, they do differ extensively in amino acid sequences, with H1 o , H1oo, and H1t being the most divergent subtypes. The different subtypes also exhibit distinct expression patterns during development, with the tissue-specific H1oo and H1t being the most extreme examples. Certain H1 subtypes can constitute a major portion of the linker histones in specific cell types. For example, H1t represents 40 -50% of the H1 in pachytene spermatocytes (8). Because linker histones are nearly equimolar with nucleosome core particles in most higher eukaryotic cells, this implies that in some cell types a large fraction of the nucleosomes contains a particular H1 subtype, e.g. H1t in developing spermatocytes. Nevertheless, gene inactivation experiments in mice have shown that elimination of any one of several of these subtypes does not disrupt normal development, even in tissues in which the specific subtypes are highly abundant (9 -14). For example, mice lacking H1t were fertile and showed no detectable defect in spermatogenesis (9,13,14). The most likely explanation for the absence of a phenotype in these single H1 null animals is that other H1 subtypes compensate for the lost subtype. Consistent with this view, analysis of chromatin from such mice showed that indeed the levels of other subtypes are increased so as to maintain a normal ratio of total H1 linker histones to nucleosome core particles (9).
Under these circumstances, one approach to investigating the in vivo roles of linker histones in mammals is to attempt to reduce the stoichiometry of H1 to nucleosome core particles by generating compound H1 null cells or mice. This is an arduous task for most H1 subtypes, because the genes for six of the eight subtypes are tightly linked on mouse chromosome 13 (7). We reported previously that in Hist1h1 tm1Ais /Hist1h1 tm1Ais (denoted H1t Ϫ/Ϫ or H1t-null) germ cells, the H1a subtype constitutes more than 50% of the remaining H1 (9). Therefore, we generated H1t and H1a null mice by two sequential genetargeting steps in mouse embryonic stem (ES) 1 cells. The double null germ cells from such mice have a 25% decrease in ratio of H1 to nucleosomes. Remarkably, the reduction in H1-tonucleosome stoichiometry did not perturb spermatogenesis or cause any apparent meiotic abnormalities, nor did it affect global gene expression. However, changes in expression of certain genes were found.

Disruption of the H1a Gene in ES Cells and Generation of Chimeric
Mice-The genomic DNA clones used in construction of the H1a targeting vector were isolated by screening a 129X1/SvJ (briefly, 129X1) genomic library as described previously (15). To generate an H1a targeting vector (Fig. 1A), a 1.2-kb SacII-TthIII fragment was inserted into the XbaI site of pTK-KS vector by blunt-end ligation as the short 5Ј-arm (Fig. 1A). The pTK-KS vector already contains a PGK-TK cassette. Then a 6.5-kb EcoRI fragment was inserted into the EcoRI site of the plasmid as the long 3Ј-arm. Finally, a PGK-Hygro cassette was bluntend-inserted into the BamHI site of the plasmid in between these two arms. To generate ES cells with a targeted disruption at the H1a locus, 25-50 g of NotI-linearized H1a targeting vector (Fig.1A) were electroporated into 2-4 ϫ 10 7 T3-12 ES cells containing a disrupted H1t allele, as described previously (9). The T3-12 ES cell line was shown previously to provide very good germ-line transmission of the disrupted H1t allele. ES cell clones resistant to 180 g/ml G418, 150 g/ml hygromycin, and 2 M gancyclovir were picked after 9 -10 days and transferred to duplicate 48-well plates. Ten ES clones containing a modified H1a allele were identified among 192 triply resistant clones analyzed by Southern blot hybridization (Fig.1B). Four clones (D18, D32, E43, and E45) were injected into C57BL/6-recipient blastocysts, and the blastocysts were transferred to CD-1 pseudo-pregnant females to generate chimeric mice as described previously (16). Mouse tail DNA was prepared as described previously (16).
Preparation and Analysis of Histones-Mice were sacrificed by cervical dislocation, and the testes were excised and immediately rinsed with Dulbecco's modified Eagle's medium. Seminiferous epithelial cells were prepared as previously described (17,18). The germ cells were isolated from testes, chromatin was prepared from the nuclei of the isolated spermatogenetic cells, and histone proteins were extracted from the chromatin with 0.2 N sulfuric acid and fractionated by high performance liquid chromatography (HPLC) as described previously (9). The effluent was monitored at 214 nm, and peak areas were determined with a Hewlett Packard peak integrator program. The percentage of each H1 subtype in the samples was calculated by measuring the ratio of each peak area to the total area of all H1 peaks after normalizing for differences in the number of amino acids in the individual H1 species. The ratio of total H1 to nucleosomes was calculated by dividing the total area of all H1 peaks by one-half the area of the H2b peak after normalizing for differences in the number of amino acids in H2b and an average sized H1 molecule.
Germ Cell Culture and Treatment-Isolated testicular germ cells were prepared as described previously (17) and washed and resuspended in HEPES-buffered minimal essential medium (Sigma) supplemented with 25 mM NaCHO 3 , 5% v/v fetal bovine serum, 20 mM sodium lactate, 50 g/ml penicillin, and 100 g/ml streptomycin (19). The cells were plated at a concentration of 2.5 ϫ 10 6 cells/ml in 4-well Nunclon dishes and were cultured at 32°C in a humidified atmosphere of 95% air/5% CO 2 . After a 2-h recovery period, the cells were treated with 5 M okadaic acid (OA) for 4 h. An equal volume of the solvent, 100% EtOH, was added to the control wells.
Air-dried Chromosome Preparations-Chromosome condensation was assessed in air-dried chromosomes prepared by a modification of the Evans procedure (20). Briefly, 2.5 ϫ 10 6 cultured cells were pooled and pelleted by centrifugation. After washing in 2.2% sodium citrate, the cells were allowed to swell in a 1% (hypotonic) sodium citrate solution. After swelling, the testicular germ cells were fixed in a 3:1 absolute ethanol/glacial acetic acid solution. The cells were dropped onto a prewarmed slide and allowed to air dry. The air-dried cell preparations were stained with Giemsa and scored for the presence of meiotic metaphase I figures or other degrees of chromosome condensation using brightfield optics with a 40ϫ objective.
Surface-spread Chromatin Preparations-Cells were fixed for indi-rect immunofluorescence analysis in 2% paraformaldehyde, 0.03% SDS as previously described (21). The antisera used were polyclonal anti-SYCP3 (22), anti-phospho-H2AX (Upstate Biotechnology), anti-histone H1t (23), and anti-phospho-histone H3 (Upstate Biotechnology). Secondary antibodies, conjugated with either rhodamine or fluorescein, were purchased from Pierce (Rockford, IL). Coverslips were mounted with Prolong Antifade (Molecular Probes, Eugene, OR) containing 0.1 g/ml 4Ј,6-diamidino-2-phenylindole to visualize the DNA. Localization was detected with a 100ϫ objective on an Olympus epifluorescence microscope. Images were captured to Adobe Photoshop with a Hamamatsu C5810 color chilled charge-coupled device camera. Histological Analysis of Mouse Testes-Testes were fixed and stained with hematoxylin and eosin and examined as described previously (9). Epididymal sperm numbers were determined by counting with a hemocytometer as described previously (9).
Microarray Procedures-A description of the microarray hardware and procedures used in this study are available from www.aecom.yu.edu/home/molgen/facilities.html. cDNA fluorescent probe preparation, prehybridization, and hybridization were performed as described (24). Briefly, for cDNA microarray analysis, RNA was extracted from germ cells of wild-type and H1a, H1t null littermate male mice using TRIzol reagent (Invitrogen). 100 g of total RNA pooled from the germ cells of three males were used in each experiment. Total RNA was reverse-transcribed in the presence of Cy3-dUTP or Cy5-dUTP with Superscript II RT. The cDNA products from wild-type and mutant cells were combined and purified on a Micron YM 50 column and hybridized to a 9000ϩ gene mouse cDNA microarray overnight at 50°C. Hybridized cDNA microarrays were washed and scanned on an Axon GenePix 4000, and signal intensities were extracted. Four replicate array hybridization experiments were carried out, with the dyes were exchanged between the samples in the experiments.
Microarray Quality Control and Data Analysis-The spot intensity was extracted from the array image using GenePix Pro 3.0 as described (24), and the intensity data file was imported into R statistical packages for data analysis. To ensure the quality of spots used in data analysis, the statistical t test between spot intensity and surrounding background intensity was carried out for each spot of each channel. The spots with p Ͼ 0.001 for both Cy3 and Cy5 channels were eliminated from further analysis. The intensities of the remaining spots were adjusted by subtracting the surrounding background intensity from the spot intensity. The ratio of the experimental sample over the control sample was calculated and further log transformed with base 2 (denoted as log 2 R/G, where R and G are the intensities of the experimental sample and the control sample, respectively), because the log-ratios form a more normal distribution than the ratios. To avoid systematic errors such as unequal loading of cDNA samples and bias of dye labeling and incorporation, intensity-dependent normalization within the array was performed to balance the intensities of both channels by applying a Lowess smooth fit function to the distribution of log-ratio (log 2 R/G) and log-intensity (1/2 log R*G) (25). The correlation coefficient of normalized log-ratios for all spots between two replicates was calculated to determine the good reproducibility of the replicates, and a cutoff for the correlation coefficient of 0.5 was used. To identify differentially regulated genes, the filtered data was imported into an Access data base and queried for genes with the median of log 2 of ratios Ն 0.58 or Յ Ϫ0.58 considered to be up-regulated (1.5-fold or more) or downregulated (1.5-fold or more), respectively, and a p value of Ͻ0.05 of statistical t test on the repeated measurements. A cutoff of 0.584 for the median of log-ratios was determined by data from a number of arrays generated by the same array platform and experimental procedure as performed in this study.
Northern Blot Analysis-Total RNA was isolated from male germ cells and analyzed by Northern blotting as described previously (26).

Generation of ES Cells with Disrupted Alleles of Both H1a
and H1t-To generate ES cells with disrupted H1a and H1t genes, we utilized a previously described ES cell clone (T3-12) in which one allele of the H1t gene had been inactivated and replaced with a PGK-Neo cassette (9). The mouse H1a gene does not contain introns, and it encodes a protein of 213 amino acids. The single short exon of the H1a gene enabled us to replace the entire H1a coding sequence with a selection marker. The H1a targeting vector (Fig. 1A) contained a PGK-Hygro cassette that allows positive selection for transfectants with hygromycin and a PGK-TK cassette that allows negative

FIG. 1. Targeted disruption of the H1a gene in mouse ES cells and mice. A, homologous recombination strategy in H1t(ϩ/Ϫ) ES cells. The
H1a targeting vector (top) was constructed by inserting a 1.3-kb SacII-TthIII fragment just upstream of the H1a start codon into the XbaI site of the pKS-TK vector, which contains a PGK-TK cassette in the pSK-Bluescript vector. Next a 6.5-kb EcoRI fragment beginning about 0.5 kb downstream of the stop codon was inserted into the EcoRI site of the vector. Finally, a blunt-ended 1.8-kb PGK-Hygro cassette was inserted into the blunted BamHI site between the two previously inserted fragments. A homologous recombination event (Xs) between the targeting vector and the endogenous H1a locus (middle) results in production of a modified H1a locus (bottom), in which a 1.2-kb fragment, including the entire H1a coding sequence along with 0.5-kb 3Ј non-coding sequence is removed. A 0.45-kb SmaI-EcoRV fragment (5Ј-probe) lying outside of the targeting construct was used as a DNA probe for Southern blotting analysis of DNA from ES cells and mice. B, identification of ES cell clones containing the modified H1a allele. DNA (10 g) was digested with EcoRI and hybridized with the 5Ј-probe shown in A. The expected positions of the hybridizing fragments from the wild-type and the modified H1a loci and their respective sizes are indicated. The 4.4-kb hybridizing band (globin transgene locus) is due to hybridization between contaminating plasmid sequences in the 5Ј-probe and multiple copies of an inactive globin transgene present in the parental WW6 ES cells (47). C, PCR genotype analysis of agouti offspring from chimeras produced with double targeted ES cells. Male chimeras produced from each of the four ES cell clones D18, D32, E43, and E45 were mated with C57BL/6 females. Genomic DNA was prepared from the agouti offspring and analyzed by PCR for the presence of either the disrupted H1t gene (H1t-Neo) or the disrupted H1a gene (H1a-Hygro). All agouti progeny derived from ES cell lines D18 and E43 were shown to be heterozygous for either the H1a-modified allele or the H1t-modified allele. All agouti progeny derived from ES cell lines D32 and E45 were shown to be wild-type for both H1a and H1t genes or heterozygous for both the H1a-modified allele and the H1t-modified allele. D, genotype analysis of offspring from parents heterozygous for either the modified H1a allele alone (H1a knockout) or both the modified H1a and H1t alleles (H1a and H1t double knockout). 10 g of tail DNA from offspring was digested with EcoRI, and Southern blots were hybridized with either the H1a 5Ј-probe (upper panel D) shown in panel A or the H1t 5Ј-probe (lower panel D) (9). Other details are as described for panel B. The deduced genotype of each mouse is indicated above each lane. selection with gancyclovir. Homologous recombination between the targeting construct and the endogenous H1a locus resulted in the generation of a modified H1a allele in which the entire H1a coding sequence was replaced by the PGK-Hygro cassette (Fig. 1A). The targeting vector was linearized by NotI digestion, electroporated into T3-12 ES cells, and cells resistant to G418, hygromycin, and gancyclovir were selected. The gancyclovirnegative selection resulted in a 4-fold enrichment of G418 and hygromycin-resistant clones. DNA from a total of 192-resistant clones was screened by both recombination-specific PCR and Southern blot hybridization with a 5Ј probe lying outside of the H1a targeting construct (Fig. 1A). Ten positive clones in which a homologous recombination event had occurred at the H1a locus were identified (examples shown in Fig. 1B). Southern blot analyses with a H1t gene-specific 5Ј probe (9) confirmed that all ten clones were H1t (ϩ/Ϫ), indicating that the disrupted H1t allele was not lost during ES cell propagation and passage (data not shown). The targeting of the H1a allele can occur in cis or in trans to the previously inactivated H1t allele. The former type of event can be used to generate H1a, H1t double mutant mice, whereas the latter type of event can be used to generate H1a mutant animals.
Generation of H1a-null Mice and H1a-and H1t-double-null Mice-To generate both H1a homozygous mutant mice and mice homozygous mutant for both H1a and H1t, four ES clones (D18, D32, E43, and E45) containing disrupted H1a and H1t alleles (Fig. 1B) were injected into C57BL/6 blastocysts to produce chimeric mice. All four cell lines gave rise to chimeras ranging in chimerism from 50% to 99% as judged by coat color. Two to four male chimeras from each line were mated with C57BL/6 females. All males mated produced agouti progeny, indicating that all four ES cell lines contributed to the germ line. Some agouti progeny from all four lines were shown to contain the disrupted H1a allele by PCR (Fig. 1C). These results were confirmed by Southern blot analysis (data not shown). By analyzing the genotypes of the agouti animals at both the H1a and H1t loci it is possible to determine whether the targeting of the H1a gene occurred in cis or trans to the inactivated H1t allele. All agouti progeny from D18 and E43 double-targeted ES cells were either heterozygous at H1a or heterozygous at H1t (Fig. 1C, left panel), indicating that the mutated H1a allele and the disrupted H1t allele were in trans in these two lines. On the other hand, all agouti progeny from the D32 and E45 cell lines were either wild-type at H1a and H1t loci or heterozygous at both loci (Fig. 1C, right panel), suggesting that the mutated H1a and H1t alleles were located on the same chromosome in these two lines.
Mice heterozygous for either the H1a mutation or both the H1a and H1t mutations were phenotypically normal. To determine the possible effects of the H1a mutation alone, mice derived from the D18 and E43 cell lines and heterozygous for the modified H1a allele were interbred, and tail DNA from the progeny was analyzed by PCR or Southern blot hybridization (Fig. 1D, left panel). Of 197 animals analyzed by either PCR or Southern blot hybridization, 43 (22%) carried only wild-type H1a alleles, 111 (56%) had one copy of the H1a modified allele, and 43 (22%) carried two copies of the H1a-mutated allele. The ratio of the three classes of mice is consistent with Mendelian transmission of the two alleles. H1a homozygous mice were normal in size and body weight (Table I), and they were indistinguishable from heterozygous and wild-type littermates. Male H1aϪ/Ϫ mice also did not exhibit any abnormalities in spermatogenesis by several criteria. They were fertile, producing litters of normal size, and both their testis weight and epididymal sperm numbers were normal (Table I). No abnormalities were detected in testis histology. These results are consistent with an earlier report (12) indicating that mutation of the mouse H1a gene does not impair normal development or spermatogenesis.
Because the entire H1a coding region was deleted in the modified H1a allele, H1aϪ/Ϫ mice would be expected to lack H1a protein. To demonstrate that homozygous mutant animals do not have H1a protein, total histone extracts of germ cell chromatin were analyzed by reverse-phase HPLC. The HPLC chromatogram of germ cell histones from wild-type animals had a peak eluting at ϳ50 min ( Fig. 2A), which is absent in germ cell histone extracts from H1aϪ/Ϫ animals (Fig. 2B). To prove that this peak is indeed H1a, we collected the material eluting in this position and subjected it to time-of-flight mass spectrometry. The mass spectrograph showed a single component with a molecular mass of 21,705 Da (data not shown), consistent with the expected size of H1a protein (21,790 Da) predicted from the H1a gene sequence after accounting for cleavage and acetylation of the resulting N-terminal serine. Because this peak is missing from germ cells extracts of H1aϪ/Ϫ mice, we conclude that the modified H1a allele is indeed a null mutation. The H1a peak was also not present in extracts of liver and spleen from the H1aϪ/Ϫ mice (data not shown).
Reduced H1 Content in Germ Cells of Mice Null for Both H1a and H1t-To generate animals null for both H1a and H1t, H1aϩ/Ϫ, H1tϩ/Ϫ mice derived from cell lines D32 and E45 (in which the mutant H1a and H1t alleles were linked) were intercrossed and tail DNA from the F2 progeny was analyzed by PCR or Southern blot hybridization (Fig. 1D, right panel). Of 171 animals analyzed by these methods, 39 (23%) carried only wild-type alleles, 88 (51%) had one copy of the mutated H1a and H1t alleles, and 44 (26%) carried two copies of the chromosome containing both the H1a and H1t null alleles. The ratio of the three classes of mice is consistent with Mendelian transmission of the doubly mutated chromosome. H1a, H1t double homozygous mutant mice were normal in weight and size, and they were indistinguishable from heterozygous and wild-type littermates (Table I). These results indicate that the presence of both the modified H1a allele and the disrupted H1t allele does not impair normal development.
Because both the H1a and H1t disruptions are null mutations, it is expected that H1aϪ/Ϫ, H1tϪ/Ϫ mice would be completely lacking both H1a and H1t proteins. This expectation was confirmed by HPLC analysis of total histone extracts of male germ cell chromatin from such mice (Fig. 2C).
To determine the effect of eliminating both H1a and H1t on the H1 subtype composition and H1-to-nucleosome stoichiometry, we analyzed these parameters by HPLC of extracts from germ cell chromatin of wild-type and homozygous double mutant male littermates. As shown in Fig. 2A, this method resolves the H1 histones in male germ cells into six peaks; H1d and H1e migrate as a single peak and therefore their amounts cannot be estimated separately unless the peak containing them is collected and subjected to mass spectrometry (27). As discussed previously (11), the HPLC analysis also allows an estimate of the amount of linker histones in cells relative to nucleosomes, by measuring the total amount of all of the H1 subtypes relative to a nucleosomal histone such as H2B. Quantitative measurements from analyses like that shown in Fig. 2, performed on germ cell chromatin from five 3-month-old wildtype and H1aϪ/Ϫ, H1tϪ/Ϫ littermates showed that elimination of both H1a and H1t caused a 25% reduction in the total H1-to-nucleosme ratio, from 0.78 in wild-type germ cells to 0.59 (Table IIb). The value for this ratio in wild-type germ cells is in good agreement with our previous measurements (9), and it is quite similar to measurements we have reported in other tissues (10,11). The observed reduction in H1 content requires elimination of both the H1t and H1a subtypes, because the ratio was not altered in H1t null mice (9) nor in H1a null mice (Table IIa). It should be noted, however, that H1 stoichiometry was not reduced in proportion to the representation of the H1a and H1t subtypes in germ cells, which is nearly 67%. Instead, other subtypes, mainly H1c, H1d, and H1e, were deposited in chromatin in increased amounts in H1a, H1t null cells, so as to partially compensate for the two missing subtypes. Mass spectrometry analysis of the H1d plus H1e peak showed that the ratio of these two subtypes is very similar in wild-type and H1a, H1t null germ cells. Thus, both H1d and H1e increased about 400% in the double null cells, followed by H1c, which increased about 250% (Table IIb). Nevertheless, these increases are not sufficient to maintain normal H1 to nucleosome stoichiometry. Characterization of Germ Cells Null for Both H1a and H1t-Having determined that elimination of both H1a and H1t leads to a reduction in the molar ratio of total H1 to nucleosomes in germ cells, we sought to examine the effects of the H1 deficiency on the properties of the cells and the properties of their chromosomes during meiosis. To determine whether H1aϪ/Ϫ, H1tϪ/Ϫ male mice are fertile, they were bred to both wild-type and corresponding null mutant females. Litter sizes from these matings were indistinguishable from those of wild-type matings, and their progeny all appeared normal. However, mice can be fertile with less than 10% of the normal number of mature sperm (28). Therefore, it was important to measure the actual number of mature sperm in the double mutants and to examine the testes for any abnormalities. No significant differ- ences were detected in testis weights of wild-type, heterozygous, or homozygous double mutant mice (Table I). To compare the number of mature sperm produced by wild-type and mutant mice, the sperm from the cauda epididymis were isolated and counted. No significant differences were detected in sperm number and sperm motility among the three classes of mice (Table I). Examination of hematoxylin-eosin-stained paraffinembedded sections of testes from 3-month-old wild-type and H1aϪ/Ϫ, H1tϪ/Ϫ littermates did not reveal any abnormal histological features in the mutant mice (data not shown). The testes of H1aϪ/Ϫ, H1tϪ/Ϫ mice contained closely packed seminiferous tubules and limited interstitial space, and the diameter of the seminiferous tubules, the thickness of the seminiferous epithelium, and the size of the lumen appeared normal. The seminiferous epithelium of H1aϪ/Ϫ or H1aϪ/Ϫ, H1tϪ/Ϫ testes contained Sertoli cells, spermatogonia, and multiple layers of spermatocytes, including pachytene spermatocytes, round spermatids, and condensed spermatids, and the lumen contained testicular sperm. Therefore, using these methods no abnormalities in the process of spermatogenesis could be detected in the testes of H1aϪ/Ϫ, H1tϪ/Ϫ mice.
H1aϪ/Ϫ, H1tϪ/Ϫ spermatocytes were further analyzed for defects in meiotic processes by using antibodies against proteins involved in synaptonemal complex formation, recombination, and chromosome condensation. To monitor homologous chromosome pairing in mutant spermatocytes, localization of the synaptonemal complex protein SYCP3 was assessed (Fig.  3). SYCP3 is a component of the lateral elements of the synaptonemal complex that is first detected in leptotene spermatocytes on the unpaired homologous chromosomes. As the spermatocytes progress through prophase I, the localization of SYCP3 changes, allowing for the identification of specific cell types. At pachytene, SYCP3 is localized to the synapsed homologous chromosome pairs with 20 distinct synaptonemal complexes visible. SYCP3 persists on the centromeres through the first meiotic division phase (29). As assessed by localization of SYCP3 (Fig. 3) and SYCP1 (data not shown) no defects in homologous chromosome pairing were detected in mutant spermatocytes and complete synapsis of the homologous chromosomes was observed (Fig. 3).
A hallmark of meiosis is the appearance of double strands breaks that occur as a result of recombination. The phosphorylation of the histone H2A variant, ␥-H2AX, has been shown to be a marker for the presence of double strand breaks that occur as a result of meiotic recombination (30). Leptotene and zygotene spermatocytes from H1aϪ/Ϫ, H1tϪ/Ϫ mice showed intense localization of ␥-H2AX over the entire nuclear area, and the same localization pattern was seen in the controls (Fig.  3, E and F). By pachynema the localization of ␥-H2AX showed the expected restriction to the sex body in both mutant and control pachytene spermatocytes (Fig. 3, G and H). The absence of staining over the autosomal chromosomes in mutant pachytene spermatocytes suggests that repair of the doublestrand breaks was not impaired. The localization of RAD51, a protein thought to participate in the repair of double strand breaks (31,32), was also the same in both mutant and wildtype controls (data not shown).
Because it is possible that spermatocyte histones regulate chromosome condensation at the time of the meiotic division, we assessed both localization of phosphorylated histone H3 (a marker of division-phase chromosome condensation) and the ability of chromosomes in cultured mutant spermatocytes to respond to precocious induction of the division phase. The localization of the phosphorylated form of histone H3 did not differ in pattern or timing between the double mutant spermatocytes and their wild-type littermate controls (Fig. 4). To monitor further the competence of the histone-depleted mutant chromosomes to condense at division phase, spermatocytes were cultured in the presence of the type I and IIa phosphatase inhibitor okadaic acid (OA), which induces a precocious metaphase-like state characterized by the condensation of chromosomes (19). In this experimental model, the double null spermatocytes were able to condense their chromosomes in a meiotic metaphase I configuration with no detectable differences from their wild-type littermate controls in the degree of condensation (Fig. 4).
A TUNEL assay was then performed on the sections to determine if the double null mice have increased levels of apoptosis. No increase in cell death was detected compared with wild-type littermate controls (data not shown).
Reducing H1 Content in Germ Cells Leads to Changes in Expression of Specific Genes-Because reducing the H1 content of germ cells did not appear to affect their development or meiotic chromosome properties, we sought to determine whether the loss of H1a and H1t affected gene expression. We compared the gene expression profiles of wild-type and H1aϪ/Ϫ, H1tϪ/Ϫ germ cells by reverse-transcribing total RNA from each source in the presence of Cy3 or Cy5 dUTP and hybridizing a mixture of the two products to microarrays containing PCR products of about 9000 cDNA clones. After processing the data according to our criteria for spot intensity relative to background and reproducibility (see "Materials and Methods"), we found that expression of only 17 genes differed between the two types of samples by more than 2-fold. Interestingly, all 17 genes were expressed at lower levels in the mutant germ cells as compared with wild-type germ cells. By lowering the cut-off to 1.5-fold changes, we found 85 more genes that differed between the two samples. Once again, genes down-regulated in the mutant cells predominated, with 67 of the 85 genes expressed at lower levels in the mutant cells. The known genes with altered expression fell into many functional groups (Table III). The changes in expression detected with the microarrays were confirmed for several genes by Northern blot hybridization with germ cell RNA from three wild-type and mutant littermate pairs (Fig. 5). One of the up-regulated genes is a putative spermatogenic-specific transcription factor (AA064110), but most of the affected genes are not enriched in germ cells. These results indicate that loss of H1a and H1t and the resulting reduction in H1 to nucleosome stoichiometry does not affect global gene expression; instead expression of specific genes is affected.

DISCUSSION
Our previous work (9) and that of other investigators (13,14) showed that the testis-specific linker histone H1t is not essential for spermatogenesis in mice. As discussed above, one possible explanation is that other H1 subtypes can compensate for loss of H1t. This explanation is supported by analysis of chro-  matin in H1t-null germ cells, which showed that other H1 subtypes are indeed increased in these cells so as to maintain a normal H1-to-nucleosome core stoichiometry (9). The H1a subtype, in particular, appeared to be involved in compensation, as it represents more than 50% of the remaining linker histone in H1t-null germ cells. The present work was undertaken with the goal of creating an actual linker histone deficiency in murine germ cells. This goal was achieved by generating two linked null alleles for H1a and H1t through a two-step, sequential gene inactivation strategy in ES cells, and then producing mice that are homozygous null for both genes. The nucleosome core particle is the fundamental organizational component of eukaryotic chromatin, and most of the DNA in cells is associated with core particles. Most eukaryotic cells also contain nearly one H1 linker histone molecule for each nucleosome core particle (1). It might be expected that alterations in the levels of these abundant cellular constituents would not be tolerated. Reductions in the cellular levels of core histones are lethal in the yeast Saccharomyces cerevisiae (33,34). Surprisingly, however, the linker histone is not needed for viability in yeast (35,36), although the normal level of this histone relative to nucleosomes appears to be significantly lower in yeast than in other eukaryotes (37). H1 also is not essential in several other unicellular eukaryotes (35, 36, 38 -40). However, we recently reported that linker histones are essential in mice (41). Elimination of the somatic linker histones subtypes, H1c, H1d, and H1e, led to mid-gestational embryonic lethality. The null embryos exhibited a 50% reduction in the level of total H1 and a broad range of phenotypic defects.
The role of linker histones in mammalian spermatogenic cells is also of interest for several reasons. These cells contain H1t, one of the two tissue-specific linker histones in mammals. During the process of spermatogenesis, there are numerous and complex changes in chromosome structure, in meiotic and post-meiotic stages. There also are marked changes in chromatin protein composition, including, ultimately, complete replacement of histones by other basic proteins (8,42). It might be expected that such cells and processes would be exquisitely sensitive to alterations in the balance of linker histones and nucleosomal histones. On the other hand, because germ cells are not essential for viability of mice, they are ideal for studying the effects of histone changes on chromatin stoichiometry.
By inactivating the genes encoding both H1a and H1t, we eliminated two H1 subtypes that normally constitute more than 65% of the total H1 histones in germ cells. However, this did not result in a 65% reduction in linker histone content in the cells. As we have observed previously, the synthetic capac- ity of the remaining H1 genes is sufficient to partially compensate for loss of the two H1 genes. In this case, H1c, H1d, and H1e constitute most of the H1 present in the double null germ cells, with H1c representing nearly 50% of the remaining H1. Nevertheless, the double gene inactivation did lead to a 25% reduction in the H1-to-nucleosome core stoichiometry (Table  II). Furthermore, we think that the actual reduction of H1 in pachytene spermatocytes may be considerably higher than 25%, because H1t represents 40 -50% of the linker histone in pachytene spermatocytes (8), whereas it is only about 30% of the H1 in germ cells, which we used for our measurements. Based on our previous studies of compound and mutant embryos null for the somatic H1s (41), one might have expected that spermatocytes with a 25% or more reduction in total H1 would have obvious defects. But we did not detect any differences in spermatogenesis between wild-type and H1a, H1t double null mice, including differences in testis weight or histology or number of mature sperm. We also did not detect any abnormalities in meiotic processes, using antibodies against various proteins involved in homologous chromosome pairing, recombination, and chromosome condensation. Interestingly, there were no observable morphological differences in meiotic chromosome synapsis or condensation in the mutant spermatocytes, implying that normal stoichiometry of H1 histones is not critical for these dynamic chromatin-remodeling processes. Our analyses also suggest that other meiotic processes, such as formation of recombination-related DNA double strand breaks and establishment of chiasmata as a result of recombination, are unaffected. Perhaps a greater reduction in H1 amount is needed to perturb these processes. The H1c gene would be an appropriate target for inactivation in the H1a, H1t doubly targeted ES cells, because H1c represents nearly 50% of the H1 remaining in double null germ cells (Table IIb).
One possible explanation for the absence of phenotypic abnormalities in the H1a, H1t null germ cells may lie in our finding that expression of relatively few genes is affected in the cells. Among about 9,000 genes assayed, we found only 17 genes that showed greater than 2-fold differences in expression between wild-type and double null germ cells. It is quite strik-ing that all 17 genes had reduced expression in the mutant. Even when gene expression changes above a 1.5-fold threshold were included, reductions in expression in the mutant predominated. These results might be viewed as surprising, because in vitro studies have suggested that H1 linker histones may function primarily in causing repression of gene transcription. However, very similar results were obtained in microarray studies of a yeast mutant lacking the Hholp linker histone (43). In this instance, only 27 of 6,216 genes assayed exhibited greater than 2-fold altered expression, and all 27 changes were reductions in expression. Thus, in both mouse germ cells and yeast, removal of linker histones does not affect the vast majority of genes and instead causes changes in expression of a small group of genes. In both studies the largest expression changes were reductions in expression in the mutants. However, in the mouse expression of some genes is derepressed with H1 loss. Thus H1 may have both positive and negative effects on gene transcription, a conclusion that is most evident in studies of two genes in a Tetrahymena strain lacking the macronuclear linker histone (44). The mechanisms by which H1 either stimulates or represses transcription of specific genes in vivo remain to be elucidated. Very likely a change in some aspect of chromatin structure is involved. We recently found that partial reductions in H1 amount lead to a global change in nucleosomal spacing (41). Such a change could affect transcription of certain genes in which control of transcription is very sensitive to positioning of specific nucleosomes. On the other hand, the nucleosome repeat length is unchanged in yeast, and Aspergillus nidulans mutants completely lacking linker histones (36,39,45) and nucleosome positions at specific loci were unchanged in the Tetrahymena mutant (46). Therefore, the mechanisms by which linker histones affect gene transcription in mammalian cells and unicellular eukaryotes may be quite different. The gene-inactivated mice reported provide a system in which to better understand the role of H1 in specific gene regulation in mammalian cells. FIG. 5. Northern blot analysis of two genes found to change expression in H1a, H1t null germ cells by cDNA microarray analysis. Panels show northern blots of total germ cell RNA from three littermate pairs of the indicated genotypes. The animals used for these experiments were different from the animals used for microarray analysis. Probes used were: A, spermatogenic-specific zinc finger transcription factor (AA064110); B, propionyl coenzyme A carboxylase homolog (W35841); and C, glyceraldehyde-3-phosphate dehydrogenase. The arrows in the figure indicate the position of 18 S ribosomal RNA in the gels before transfer (not shown).