We report here the solution structure, dynamics, and biochemical evidence for a functional surface of the BRCT domain from human polymerase μ This represents a first step toward understanding how these polymerases interact with the NHEJ machinery.
Structural Comparison to the BRCT Domain from TdT
Two other members of the pol X family of polymerases, TdT and pol λ
also have N-terminal BRCT domains. Pol μ
is most similar to TdT, with 41% sequence identity (2
) overall and 39% sequence identity over the BRCT domain (residues V30-E124, see ). In contrast, the sequence identity with the pol μ
BRCT domain is only 19% over the structured region of the domain (). The residues that are identical between all three domains are shown in red and highlighted in yellow in .
Recently, an NMR solution structure of the BRCT domain of human TdT has been deposited in the Protein Data Bank (PDB ID 2COE, Nagashima, et al., unpublished). The pol μ and TdT BRCT domains have the same folding topology, and the lowest energy TdT structure superimposes on the average energy-minimized pol μ structure with a backbone rmsd of 2.1 Å (). Overall, the structures are very similar, although α2 is shorter in the TdT domain compared to the μ domain. Since α2 is shorter in TdT, the position of L4, connecting α2 and β4, in the two domains is different: In the pol μ ensemble of structures, L4 appears more disordered (), and residues A71, C75, and T76 in L4 exhibit increased backbone dynamics (). L4 is also a region of low sequence conservation ().
The positions of many of the hydrophobic and aromatic residues in the two domains are very similar, such as interactions between Y33 and W82 and W104 and H68 ( and corresponding residues in 6C). Both BRCT domains have exposed hydrophobic residues on α1 and α2; in pol μ the side chains of F46, L50, and V80 are exposed; in TdT the side chains of F40, L44, and L74 are exposed (). Note that a leucine at position 40 of the TdT structure is inconsistent with a phenylalanine found at this site in both the published human TdT sequence as well as TdT from other species (see, e.g., ), suggesting that leucine is either a rare polymorphism or a mutation that arose during cloning. The solvent exposure of hydrophobic residues F46, L50, and V80 (and corresponding residues in TdT) suggested that these residues might mediate interactions with other proteins in the NHEJ repair complex.
The electrostatic surfaces of the pol μ and TdT BRCT domains are also very similar (). Both domains have a ridge of positively charged residues on one face of the protein (compare ), and large negative regions on the opposite faces of the proteins (compare ). In the pol μ domain, the positive ridge is composed of residues R44, R52, R85, and R86, whereas in TdT the ridge is composed of residues R47, R46, R38, and K31. It is conceivable that this ridge may be an interaction site for phosphopeptide binding. In the pol μ domain, the ridge is partially obscured by the acidic side chains of residues E36 and D60, which form a salt bridge with R44. However, in TdT the acidic side chains of E30 and E55 are in close proximity to R34, corresponding to R44 in pol μ, and may also form salt bridges under some conditions.
Figure 7 Electrostatic surfaces of the BRCT domains of pol μ and TdT. (A) A ridge of positively charged residues on the surface of the BRCT domain of pol μ may be the site of phosphopeptide or DNA binding. (B) Electrostatic surface of the BRCT (more ...)
Comparison to Other BRCT Domains
One of the interesting aspects of the BRCT fold is that it appears to be multifunctional. The fold accommodates direct binding of both proteins (phosphorylated and nonphosphorylated) and DNA, and it utilizes a number of different surfaces for different interactions. A variety of structures of BRCT domains have been determined (65
), although the majority of these are of tandem repeat BRCT domains, such as in BRCA1 (62
), 53BP1 (14
), and MDC1 (66
). Tandem repeat BRCTs, in particular, have been shown to bind phospho-peptides (22
). In addition, other BRCT domains appear to homo- or heterodimerize (68
). The pol μ
BRCT structure is atypical in that it appears to be the first reported (in the literature) BRCT domain structure that is neither part of a tandem grouping nor forms a stable dimer, as demonstrated here for μ
-BRCT by 15
N relaxation and light scattering. (It should be mentioned that while other BRCT domains appear in the PDB as monomeric structures (see below), without additional measurements, the oligomeric state of these domains remains uncertain).
To assess the degree of structural similarity between pol μ BRCT and individual BRCTs from tandem or dimerized domains, we carried out a number of backbone superimpositions. Domains found either in tandem pairs or as noncovalent dimers from BRCA1, XRCC1, 53BP1, MDC1, and DNA ligase IIIα all superimposed to the energy-minimized μ-BRCT domain with RMSDs of 5 Å or greater. The sole exception was the C-terminal BRCT domain from the NMR structure of BRCA1 (1OQA), which has an rmsd of 2 Å if helix α2 is excluded from the superposition; however, this BRCT domain is structurally different because the α2 helix from this BRCT domain has only ~1 turn. An example of the different orientation of helix α2 in the μ-BRCT domain vs the C-terminal BRCT domain from XRCC1 (1CDZ) is shown in Supporting Information (Figure S4).
Next, we sought to determine the extent of structural similarity between the pol μ BRCT domain (and hence pol TdT) and BRCT domains in the protein data bank that are not from tandem pairs or found as dimers. The apparently monomeric BRCT domains found in the PDB include the NAD+-dependent DNA ligase BRCT domain from T. thermophilus (PDB ID: 1L7B), the second BRCT domain of epithelial transforming growth factor 2 (PDB ID: 2COU), and the BRCT domain of poly(ADP-ribose) polymerase-1 (PDB ID: 2COK), which superimpose with backbone RMSDs of 4.8, 6.3, and 11 Å, respectively. In addition, 2COU, as with 1OQA (BRCA1) and 1IMO (DNA ligase III), lacks most or all of the α2 helix. In summary, it appears that the pol μ and TDT BRCT domains have unique structural properties, as indicated by high rmsd values with other BRCT domains, and that an element of major structural difference across all BRCT domains is the α2 helix.
Relaxation/Dynamics and Monomeric State of pol μ BRCT
We have performed a standard 15
N relaxation analysis of the backbone dynamics of human polymerase μ
. Only two BRCT domains have been previously characterized by 15
N relaxation: the C-terminal BRCT domain from BRCA1 (62
) and the BRCT domain from DNA ligase IIIα (17
). Because BRCA1 contains a tandem repeat and DNA ligase IIIα is reported to be a dimer in solution (17
), this study represents the first dynamics characterization of a singly occurring (nontandem), nondimeric BRCT domain. The correlation time for tumbling, τm
, was determined to be 10.5 ns at 10 °C. We believe that this is consistent with 15
N relaxation data on the dimeric BRCT domain from DNA ligase IIIα, even though that domain was reported to have a global τm
value of 6.82 ns at 15 °C (16
). In that same report (16
), the average R2
values were 19.5 s−1
at 15 °C, consistent with a tumbling correlation time of at least 14 ns. This is reflected in the local τm
times reported in of that paper, which are inconsistent with the global tumbling time cited in the text (16
). These considerations lead us to conclude that the pol μ
BRCT is monomeric under the conditions studied here. This has been confirmed by light scattering experiments (Supporting Information, Figure S2). For the pol μ
BRCT domain, the only region found to have increased flexibility on the ps-ns time scale is L4 (). On the other hand, slower motion on the μ
s-ms time scale occurs in loops LI and L2.
Functional Hot Spots in pol μ BRCT Are Proximal to Phospho-Binding Sites in Other BRCT Domains
Inspection of the solution structure and dynamics of pol μ BRCT allowed identification of the L1-α1 region as a surface containing residues that are potentially important for function. Specifically, the hydrophobic side chains F46 and L50 in α1 are conspicuously exposed to solvent and show high conservation in pol μ and TdT. In addition, R43, also in α1, shows evidence for flexibility on the μs-ms time scale, along with residues in L1 () that lead into α1. Because of their spatial proximity and definition of a contiguous surface (), residues R43, F46, and L50 were mutated to test their importance in NHEJ complex formation and activity using gel shift assays. Since all of these residues are found to be solvent-exposed in the pol μ BRCT structure, it was assumed that substitution of any of these residues with alanine would not result in significant structural changes. These general conclusions were supported by the 1H spectra of the mutated proteins, which were all consistent with the expectation of a conserved structure for each of the mutants tested. The assays showed that mutation of any one of these residues to alanine in the context of full length pol μ largely abolishes pol μ’s ability to form a stable complex with Ku and XRCC4-ligase IV assembled on DNA ends. This in turn strongly reduces the ability of pol μ to help Ku and XRCC4-ligase IV join ends that require prior gap filling before they can be joined. The results of these experiments demonstrate that the N-terminal region of helix α1, and possibly L1, have a pivotal role in promoting interactions with the NHEJ machinery that are critical for pol μ’s biological role.
Interestingly, this general region has been implicated in other BRCT-protein interactions. In the BRCA1–BACH1 complex, S1655 and G1656, in L1 of the N-terminal BRCT domain, form hydrogen bonds to the phosphoserine (26
). Nearly identical interactions are formed in the MDC1–H2AX complex (69
). These residues correspond to μ
-BRCT residues V35 and E36, in L1, which experience μ
s-ms fluctuations (). Several lines of reasoning suggest that the details of functional interactions in μ
-BRCT may be different than in the BRCA1–BACH1 or MDC1–H2AX complexes. First, μ
-BRCT is not a tandem domain, although it is possible that it may form heterodimers with other BRCT domains, such as that in XRCC4. Second, L1 in μ
-BRCT is composed of nearly twice as many residues as in the N-terminal BRCT domains of BRCA-1 and MDC1. Third, in the N-terminal region of α1, there are two conserved arginines (43 and 44). R44 is involved in a salt bridge to the side chain of E36, which resides in the flexible L1. These two arginines are conserved throughout species of μ
and TdT polymerases (), whereas in the N-terminal BRCA1 domain there are no positively charged side chains in this region and a glutamate at the position corresponding to R43 in μ
-BRCT. It is worth mentioning, however, that in one of the BRCA1–BACH1 crystal structures (1T29), the carboxyl group of E1661 (equivalent to R43 in μ
-BRCT) appears to be involved in an electrostatic interaction with the side chain of R
3 (−3 position) in the BACH1 peptide. This is consistent with a role for the conserved arginines in μ
-BRCT in peptide/protein (or even DNA) recognition.
In summary, the prominent solvent exposure of hydrophobic residues and evidence for microsecond-millisecond motions, along with sequence conservation information, has led to the biochemical identification of a group of spatially proximate residues likely to be important for protein interactions of the pol μ BRCT domain. The significance of these residues received strong support from assays performed on mutated analogs of pol μ. In addition, a structural comparison with the BRCT domain of TdT suggests that this interaction surface is well conserved, supporting the hypothesis that these polymerases can both be recruited by the NHEJ repair complex.