We have used a combination of approaches and technologies to unravel the physiological function of the testis-specific paralog of CTCF, called CTCFL or BORIS. We find that CTCFL is only expressed in late spermatogonia and preleptotene spermatocytes, and that CTCFL-deficient mice have defects in spermatogenesis. CTCFL and CTCF are functionally different proteins. CTCFL therefore has a unique role in the adult testis. It has been proposed that CTCFL is involved in genomic imprinting of the Igf2-H19 locus and other sites [19, 24]. However, imprint-related mutations often have embryonic phenotypes . We did not observe this in Ctcfl
mice, and despite their reduced fertility Ctcfl
mice could be bred through multiple generations. Furthermore, we have not been able to detect DNA methylation aberrations in specific loci in Ctcfl
mutant mice and in CTCFL-over-expressing cells (data not shown). This makes a role for CTCFL in DNA methylation-dependent genomic imprinting unlikely. The combined microarray data from CTCFL-deficient testis and CTCFL-expressing ES cells, and the preference of CTCFL for promoters instead suggest a function as a transcriptional regulator, required for the proper expression of a subset of male germ cell genes.
The most prominent CTCFL-binding sites in ES cells are on the promoters of the testis-specific Stra8 and Prss50 genes. The expression of these genes, and of Gal3st1, is upregulated in ES cells expressing CTCFL. Conversely, expression of Prss50 and Gal3st1 is downregulated in germ cells lacking CTCFL, at all ages examined, whereas Stra8 expression is affected at some but not all ages (data not shown). We speculate that the combined transcriptional deregulation of genes causes the testicular degeneration and reduction in fertility in Ctcfl knockout mice. Note that the expression of these genes is not completely hampered, which explains why the testicular phenotype in the knockouts is milder than the fully sterile phenotype described, for example, for STRA8- and GAL3ST1-deficient mice [29, 30, 35].
The phenotype of the Ctcfl
mice reported here only partly matches a recent report on another strain of CTCFL-deficient mice, in which exons 1 to 8 of Ctcfl were also deleted . For example, the effect of a Ctcfl deletion on the average testicular size and on Gal3st1 and Prss50 expression is similar. However, our analysis also reveals a reduction in fertility in the Ctcfl
mice not noted previously . In addition, the fact that some Ctcfl
mice have normal testis size and others have a combination of normal and abnormal seminiferous tubules was also not described. This is relevant, as this incomplete penetrance of the Ctcfl phenotype, even within a single testis, suggests that a stochastic mechanism determines whether CTCFL-deficient tubules degenerate or not. Finally, CTCFL was proposed to be present in round spermatids and to function during meiosis based on mRNA expression data . By contrast, our data show that CTCFL is expressed earlier, just prior to the onset of meiosis, and we conclude that CTCFL protein expression precedes the developmental germ cell stages that show the major phenotypes in Ctcfl knockout mice. We propose that in the absence of CTCFL, epigenetic marks controlled by this protein gradually break down in a stochastic manner. Spermatogonia and primary spermatocytes exist in syncitia, in which each cell is connected with the other cells at the same step of development via intercellular bridges. Only in syncytia where the expression of CTCFL-controlled genes has been affected beyond a specific threshold will degeneration become apparent.
Neither CTCFL nor CTCF is saturating all consensus-binding sites present in the genome, and thus the DNA sequence is not the sole determinant of CTCF(L) binding. DNA methylation and hydroxymethylation are not a decisive aspect, as comparisons of DNA (hydroxy)methylation data sets to our CTCF(L)-binding sites does not provide an explanation for why CTCFL and CTCF occupy different binding sites (data not shown) . Instead, the data suggest that binding of CTCFL and of the “master weaver” CTCF is specified by nucleosome occupancy and composition. We find that CTCFL prefers CTCF consensus sites in promoters that are embedded in regions that appear to be nucleosome-free. By contrast, CTCF is enriched on distinct sites, which are devoid of histone H3 on the binding site itself, but which are surrounded by ordered, or “phased,” nucleosomes. This preference of CTCF has already been described [11–13].
It has recently been shown that unstable nucleosomes are lost when histones are prepared with conventional conditions; thus, regions containing these histones appear as nucleosome-free in the analysis, but are in reality not free . Nucleosomes containing the variant histone H3.3 are quite unstable, and those containing both H3.3 and H2A.Z even less . Since we find a correlation between CTCFL binding and H3.3 occupancy in ES cells, H3.3 and H3.3/H2A.Z might be determinant factors able to attract CTCFL and evict CTCF. It is important to realize that in ES cells H3.3-enriched genomic regions do not require CTCFL to be set up, yet the protein prefers such areas after its induction. A similar situation may exist in testis, i.e., specific H3.3/H2A.Z-containing regions might be set up during early phases of spermatogenesis; upon its expression, CTCFL “lands” on these regions, possibly evicting CTCF from some promoters. Notably, during male meiosis, and thus subsequent to CTCFL expression, H3.3 is incorporated into unsynapsed chromatin, which is transcriptionally inactive . The function of CTCFL might be to ensure the expression status of genes by distinguishing specific promoter-associated H3.3 domains from whole chromosome domains that also contain H3.3. Through its interaction with SET1A , CTCFL might enhance H3K4 trimethylation at a subset of its binding sites.
The cohesin complex has a role in chromosome segregation, DNA-damage repair and gene regulation . Although cohesin does not have a typical DNA-binding motif, it was shown to bind primarily to CTCF consensus sites [16, 17, 51]. Moreover, the SA2 subunit of cohesin directly interacts with the C-terminus of CTCF .
Cohesin’s role in gene regulation therefore seems tied to that of CTCF. Recent studies revealed that also in ES cells cohesin binding largely overlaps with that of CTCF; however, there are ~2,000 cohesin sites with a CTCF motif that do not bind CTCF, while ~270 other cohesin sites do not have a CTCF consensus site . Our data suggest that CTCFL binds these ~2,000 cohesin sites in CTCFL-GFP-V5-expressing ES cells.
However, in normal ES cells CTCFL is not expressed, raising the questions how a specific nucleosome composition and occupancy can be built around CTCF consensus sites that appear not to be occupied by CTCF, and how cohesin can stably bind these very same sites. We hypothesize that these sites might be bound by a modified form of CTCF, such as poly(ADP-ribosyl)ated CTCF . This protein would not be able to bind DNA tightly and could be replaced very efficiently by CTCFL. Perhaps another molecular function of CTCFL in the testes is to interfere with and/or change the dynamics of CTCF and cohesin-mediated chromatin looping.
We observed competition between CTCF and CTCFL in ES cells, but only on a small subset of all CTCF-binding sites. Nucleosome occupancy and composition, CTCF(L) expression levels and posttranslational modifications on CTCF(L) could all determine whether competition between the proteins occurs on a given site. Our data reveal that CTCF and CTCFL co-localize within the nuclei of late spermatogonia and preleptotene spermatocytes, and the proteins might therefore also compete in vivo. ChIP experiments in testis extracts indeed reveal preferential binding of CTCFL at the Stra8 and Prss50 promoters and exclusive binding of CTCF to the Vps18 site. These data are consistent with binding profiles in ES cells. If competition on the Stra8 and Prss50 genes does occur in vivo, then CTCFL could be a gene activator by preventing the binding of CTCF. In Ctcfl knockout mice binding of CTCF to these genes might actually diminish their expression. However, CTCF is ubiquitously expressed in the testis, whereas CTCFL is only transiently present in spermatogonia and preleptotene germ cells. One would expect to see significant binding of CTCF to the Stra8 and Prss50 sites in the testicular extracts that we used, since most cells in these extracts contain CTCF and not CTCFL. The questions why CTCF is not highly enriched on the Stra8 and Prss50 promoters in testis, and whether these proteins compete in vivo can only be answered once there are tools available to isolate CTCFL-positive and -negative cell populations from testis so that genome-wide analyses can be performed on purified testicular fractions.
In human germ cell tumors, CTCFL is specifically upregulated in spermatocytic seminomas, which are benign testicular tumors originating from a spermatogonium or primary spermatocyte . This fits with our observed cellular localization of CTCFL and could potentially point to an oncogenic role for CTCFL in these tumors. In fact, CTCFL belongs to the group of cancer testis antigens (CTAs), genes that are normally expressed in testis yet aberrantly expressed in a variety of cancers. One model holds that competition between CTCF and CTCFL plays a role in tumorigenesis, i.e., aberrant CTCFL expression would displace CTCF, and affect DNA methylation and the expression of other CTAs, including the NY-ESO-1 and MAGE-A1 genes [22, 23], and even other important genes, such as the TERT gene, which encodes telomerase . However, while there might be a relationship between DNA demethylation and the expression of CTAs , recent reports have shown that expression of CTCFL alone is not sufficient to induce expression of CTAs [27, 57]. Furthermore, our data in CTCFL- deficient testis indicate that, if anything, CTCFL represses the Tert gene instead of activating it. To address a potential role of CTCFL in cancer, a correlation analysis of CTCFL binding, nucleosome occupancy and composition, and CTA expression in different types of cancers might be more revealing.