The constitutive insulator activity of the Fab-7
boundary is generated by a series of functionally redundant sub-elements whose activity is developmentally restricted. Here we report the identification and characterization of a factor, Elba, which confers the insulator activity of one of these sub-elements in early embryos. Elba is an unusual hetero-tripartite complex. It consists of two ~40 kDa proteins, Elba1 and Elba2, that have a C-terminal 90 amino acid BEN domain embedded in a larger 130 amino acid region of homology. The BEN domain is found in insect and vertebrate nuclear proteins and has been implicated in protein:protein interactions and transcriptional regulation (Abhiman et al., 2008
). The third protein, Elba3 has no distinctive domains and seems to be limited to the Drosophilids. All three proteins are present in the Elba complex detected in nuclear extracts and are required to reconstitute Elba DNA binding in vitro.
Our functional studies indicate that DNA binding is mediated by the C-terminal domains of Elba1 and Elba2; however, in order to form a DNA binding ‘pocket’ that recognizes the asymmetric Elba sequence, CCAATAAG, Elba1 and Elba2 must be linked in a ‘heterodimer’ (). In the tripartite Elba complex, this seems to be the role of Elba3. It brings Elba1 and Elba2 together by interacting with sequences in the N-terminal half of the two proteins. These conclusions are supported by a number of findings. First, it is possible to circumvent the requirement for Elba3 by fusing a heterologous dimerization domain, in this case GST, to the N-terminus of full length Elba1 and Elba2. When co-translated, these two fusion proteins shift the Elba probe without Elba3. On the other hand, they can still interact with Elba3 as the shift generated by the two proteins can be supershifted by the addition of Elba3. Second, the Elba1C and Elba2C proteins, which span the 130 amino acid C-terminal homology region, can reconstitute full DNA binding activity when fused to GST. Since GST is known to dimerize, but not form higher order complexes, this would argue that the Elba DNA binding pocket is likely formed by a GST-Elba1C:GST-Elba2C dimer rather than by some other more complicated combination of the two proteins. Moreover, this binary complex has the same sequence specificity as the native hetero-tripartite complex. On the other hand, the GST-Elba1C:GST-Elba2C proteins differ from the full length GST fusions in that they don't interact with Elba3.
While these findings are consistent with the model for the Elba complex shown in , there are several unresolved but important issues. For one, though our experiments indicate that the BEN domain is essential for DNA binding, we were unable to demonstrate that it is sufficient. It is possible that steric hindrance from the closely linked GST moiety prevents formation of the DNA binding pocket; alternatively, the extended homology region may contain elements that are important for DNA binding. Another apparent anomaly is that the homodimers formed by the full length or the C-terminal GST fusions don't appear to bind to the Elba probe. Since we've found that the Elba factor can bind to variants of the Elba sequence, albeit with reduced affinity, we would have expected that the homodimers would exhibit at least some evidence of DNA binding activity. Further studies will be required to resolve these issues.
As would be the case for many other boundaries, the presence of functionally redundant elements in the full length Fab-7 precludes a direct demonstration that a single factor like Elba is needed for Fab-7 insulating activity either in the context of BX-C or in transgenes assays. However, several lines of evidence argue that Elba does in fact have such a function for the endogenous Fab-7 boundary. To begin with, previous studies showed that the insulating activity of the 236-bp pHS1 Fab-7 sub-element in early embryos is compromised when the Elba recognition sequence is mutated. Moreover, when multimerized, the Elba sequence is sufficient on its own to confer insulating activity. That the Elba factor is responsible for this insulating activity is supported by the effects of RNAi knockdowns in embryos carrying either the pHS1×4 transgene or the Elba sequence multimer. For all three Elba proteins, RNAi knockdowns compromises the insulating activity of both pHS1×4 and the Elba multimer. In contrast, knockdowns of the closely related BEN domain protein, Insv, have no effect.
Our findings would also explain why the insulating activity of the Elba factor/Elba sequence is developmentally restricted. Two of the three Elba proteins, Elba1 and Elba3, are encoded by genes that are active during the mid-blastula transition, but not later in development. Since the hetero-tripartite complex is required to reconstitute DNA binding activity, Elba insulating activity would be expected to peak during the blastoderm/early gastrula stage when high levels of elba1 and elba3 mRNA are expressed. However, it should gradually dissipate after transcription of elba1 and elba3 ceases. While we don't know precisely when Elba1 and Elba3 disappear, there is little Elba DNA binding activity in nuclear extracts of 6–12 hr embryos. Moreover, even though the elba2 gene is expressed throughout development, it is not found associated with Fab-7 in 9–12 hr embryos. Importantly, the results of the EMSA and ChIP experiments provide strong support for the idea that the requirement for Elba activity evident in the RNAi knockdowns reflects a direct role in insulating activity rather than indirect role.
One important question is why does a constitutive insulator like Fab-7
utilized developmentally limited factors like Elba? A plausible explanation comes from the finding that boundaries are not autonomous entities, but rather function in combination (likely pairwise) with other boundaries (Cai and Shen, 2001
; Muravyova et al., 2001
). Both boundary competition (Gohl et al., 2011
) and boundary bypass experiments (Kyrchanova et al., 2008a
; Maksimenko et al., 2008
) indicate that some combinations are functional while others are not. Moreover, how different boundary combinations work together depends upon developmental stage and tissue (Gohl et al., 2011
). Thus, the use of stage specific factors at Fab-7
could reflect a need to optimize combinations with other nearby BX-C boundaries or the Abdominal-B promoter (Kyrchanova et al., 2008b
). A number of findings are consistent with this idea. When heterologous boundaries are used to replace Fab-7
in BX-C, they are unable to provide bypass activity, while their insulating activity can be lost in a stage and tissue specific fashion (Hogga et al., 2001
). In contrast, Fab-7
is able to replace Fab-8
(Iampietro et al., 2008
). Likewise, the boundary activity of Fab-7
in a foreign environment is not only context dependent, but also varies during development (Gohl et al., 2011
). Probably the most direct evidence that boundaries in BX-C like Fab-7
are closely matched with their flanking neighbors comes from the boundary bypass experiments of Kyrchanova et al. (2011)
. They found that bypass is observed when Fab-7
is combined with the neighboring boundaries Fab-6
, while bypass interactions with boundaries from elsewhere in BX-C are only weak at best. Interestingly, in the cases that have been analyzed most thoroughly, boundary bypass is mediated by homologous protein:protein interactions (Kyrchanova et al., 2008a
). Thus, bypass is observed when multimerized binding sites for the dCTCF boundary factor are paired, but not when dCTCF binding sites are paired with sites for example Zw5. Though Fab-8
differs from Fab-7
in that it utilizes dCTCF (Holohan et al., 2007
; Kyrchanova et al., 2011
), EMSA experiments with 0–6 hr and 6–12 hr nuclear extracts indicate that they also have binding sites for several of the same stage specific factors (Aoki et al., 2008
; Wolle et al., unpublished data). These include sites for the Elba factor, and sites for two different late (6–12 hr) stage specific factors. Conceivably this could also be true for Fab-6
Another question is the fate of Elba2 after Elba1 and Elba3 disappear. While Elba2 can't bind to the Elba sequence on its own, it could function as an insulator in other contexts using different protein partners and presumably also recognition sequences. A plausible partner would be the Notch
signaling pathway protein Insv, which is encoded by a tightly linked gene that has a similar expression pattern to elba2
. Insv and its closest mammalian relatives, the NAC proteins (which are thought to function as stem cell pluripotency genes, as well as in cocaine addiction and cancer), are believed to regulate transcription by acting as co-repressors (Cha et al., 1997
; Mackler et al., 2000
; Wang et al., 2006
; Korutla et al., 2009
; Duan et al., 2011
). However, since our findings indicate that the BEN domains of Elba1 and Elba2 are essential for DNA binding, it would be reasonable to suppose that Insv as well as the mammalian counterparts are not co-repressors, but are rather sequence specific DNA binding proteins. Interestingly, like members of the BEN domain protein family, CTCF was initially thought to be a transcriptional repressor (Filippova et al., 1996
). Thus, a seemingly plausible speculation would be that other members of the BEN family besides the two Elba proteins function as insulators restricting the action of nearby enhancer elements instead of directly repressing transcription. Potentially supporting this idea is the finding that one of the mammalian BEN domain proteins, SMAR1, is a component of the nuclear matrix and associates with Matrix Attachment Regions (MARS) (Chattopadhyay et al., 2000
). If this were the case, this would mean that CTCF is not the only vertebrate insulator protein. Also as is case for the Elba complex, modulating the expression or activity of Elba2, Insv or NAC1 in different tissues or cell types could change the regulatory domain landscapes, and potentially alter global patterns of gene expression.