Article Cite This: Acc. Chem. Res. 2018, 51, 526−533
pubs.acs.org/accounts
DNA’s Encounter with Ultraviolet Light: An Instinct for SelfPreservation? Adam Barlev† and Dipankar Sen*,†,‡ †
Department of Chemistry and ‡Department of Molecular Biology and Biochemistry, Simon Fraser University, Burnaby, British Columbia V5A 1S6, Canada CONSPECTUS: Photochemical modification is the major class of environmental damage suffered by DNA, the genetic material of all free-living organisms. Photolyases are enzymes that carry out direct photochemical repair (photoreactivation) of covalent pyrimidine dimers formed in DNA from exposure to ultraviolet light. The discovery of catalytic RNAs in the 1980s led to the “RNA world hypothesis”, which posits that early in evolution RNA or a similar polymer served both genetic and catalytic functions. Intrigued by the RNA world hypothesis, we set out to test whether a catalytic RNA (or a surrogate, a catalytic DNA) with photolyase activity could be contemplated. In vitro selection from a random-sequence DNA pool yielded two DNA enzymes (DNAzymes): Sero1C, which requires serotonin as an obligate cofactor, and UV1C, which is cofactor-independent and optimally uses light of 300−310 nm wavelength to repair cyclobutane thymine dimers within a gapped DNA substrate. Both Sero1C and UV1C show multiple turnover kinetics, and UV1C repairs its substrate with a quantum yield of ∼0.05, on the same order as the quantum yields of certain classes of photolyase enzymes. Intensive study of UV1C has revealed that its catalytic core consists of a guanine quadruplex (Gquadruplex) positioned proximally to the bound substrate’s thymine dimer. We hypothesize that electron transfer from photoexcited guanines within UV1C’s G-quadruplex is responsible for substrate photoreactivation, analogous to electron transfer to pyrimidine dimers within a DNA substrate from photoexcited flavin cofactors located within natural photolyase enzymes. Though the analogy to evolution is necessarily limited, a comparison of the properties of UV1C and Sero1C, which arose out of the same in vitro selection experiment, reveals that although the two DNAzymes comparably accelerate the rate of thymine dimer repair, Sero1C has a substantially broader substrate repertoire, as it can repair many more kinds of pyrimidine dimers than UV1C. Therefore, the co-opting of an amino acid-like cofactor by a nucleic acid enzyme in this case contributes functional versatility rather than a greater rate enhancement. In recent work on UV1C, we have succeeded in shifting its action spectrum from the UVB into the blue region of the spectrum and determined that although it catalyzes both repair and de novo formation of thymine dimers, UV1C is primarily a catalyst for thymine dimer repair. Our work on photolyase DNAzymes has stimulated broader questions about whether analogous, purely nucleotide-based photoreactivation also occurs in double-helical DNA, the dominant form of DNA in living cells. Recently, a number of different groups have reported that this kind of repair is indeed operational in DNA duplexes, i.e., that there exist nucleotide sequences that actively protect, by way of photoreactivation (rather than by simply preventing their formation), pyrimidine dimers located proximal to them. Nucleotide-based photoreactivation thus appears to be a salient, if unanticipated, property of DNA and RNA. The phenomenon also offers pointers in the direction of how in primordial evolutionin an RNA worldearly nucleic acids may have protected themselves from structural and functional damage wrought by ultraviolet light.
D
light-utilizing proteins that have been refined by billions of years of evolution.2 Recent studies on certain photochemical properties of DNAs, however, suggest that the cell’s nucleic acids (both DNA and RNA) inherently manifest certain photolyase-like capabilities, effectively in-built mechanisms for self-protection from ultraviolet light.3 In the cell, most classes of DNA chemical damage occur stochastically at a low rate throughout the genome. By contrast, pyrimidine dimers form with high efficiency upon exposure to UV light. 4 Hotspots for UV damage tend to locate predominantly in repetitive elements, with little bias between intron or exon sequences;5 however, certain archea that face
NA is the polymer used by all free-living life forms as a long-term repository of their genetic information. Damage to DNA, whether from exogenous or endogenous sources, can determine the survival or death of the organism. Light is the ultimate source of metabolic energy that fuels the biosphere, but certain wavelengths of light are known to lead to a pernicious form of DNA damage, the pyrimidine dimer. In the clinic, absorption of UV light in the 200−300 nm range by a bacterial pathogen’s DNA may be used as a potential disinfectant, but paradoxically, subsequent irradiation with benign visible light may reverse the effect.1 This is due to the presence within bacteria of photolyase enzymes, visible-lightutilizing proteins that can perform direct photochemical repair of thymine dimers. A variety of photolyases have now been characterized in exquisite detail; these are highly sophisticated © 2018 American Chemical Society
Received: November 21, 2017 Published: February 8, 2018 526
DOI: 10.1021/acs.accounts.7b00582 Acc. Chem. Res. 2018, 51, 526−533
Article
Accounts of Chemical Research
enhancements of up to ∼380-fold relative to the unsensitized CPD photoreactivation reaction. Kinetic isotope effects measured on this antibody system indicated that as with the cofactor-denuded photolyase, charge transfer occurred from a photoexcited tryptophan within the antibody’s site of combination with the bound CPD.18 Catalytic antibodies constitute one category of modestly evolved protein enzymes; however, we have been interested in discovering novel catalysts made from nucleic acids. The catalytic potential of nucleic acids has significant implications for the origin of life. The renowned “RNA world hypothesis” posits that at very early stages in the development of life, RNA (or an RNA-like polymer) served both catalytic and genetic functions.19 Indeed, no alternative theory satisfactorily explains the multitude of roles RNA and nucleosides continue to play in contemporary biochemistry. Given the demonstration that an abzyme is able to photoreactivate CPDs via charge transfer from its tryptophan, it appeared reasonable that a catalytic RNA/ribozyme (or its surrogate, a catalytic DNA/DNAzyme) could, with the help of a tryptophan-like cofactor, achieve a comparable photoreactivation. In 2003, we set out to explore this possibility, electing to use the tryptophan metabolite serotonin, which has both superior solubility and diminished spectral overlap with DNA relative to tryptophan itself.16 In vitro selection (“SELEX”) out of large random-sequence RNA or DNA libraries is a powerful methodology for the de novo identification of ligand-binding and/or catalytic RNA or DNA sequences.20 Our goal was to select a photocatalytic DNAzyme capable of repairing a CPD within a DNA context in a serotonin-dependent manner.21 Key to this selection design was the synthesis of a suitable CPD-containing DNA substrate, which was generated by aligning two short oligonucleotides, T1 and T2, into a double helix such that terminal thymines from T1 and T2 were juxtaposed in the duplex. Addition of a triplet sensitizer, acetophenone, and irradiation with 365 nm light ensured that only the desired cis-syn CPD formed between the juxtaposed thymines, covalently joining T1 and T2. The resulting composite oligonucleotide, TDP, lacked a phosphodiester between its dimerized deoxythymidines. It was then used as a PCR primer to amplify a 95-nucleotide DNA library that included a stretch of 40 random nucleotides. The resulting duplex DNA library was treated to recover its component single strand that incorporated the gapped CPD, folded in salt, with the expectation that individual sequences within the library that showed preferential photoreactivation would suffer concomitant loss of T1. This would lead to a significant electrophoretic mobility shift of the photoreactivated DNA relative to unreactivated library members. Accordingly, this library was incubated in a buffer solution with added serotonin and irradiated with UV light (λ ≥ 300 nm). Library members undergoing rapid photoreactivation were recovered and amplified using PCR.21 Curiously, the above SELEX experiment yielded two classes of DNA sequences. As expected, a class of serotonin-dependent DNAzymes was obtained (with action spectra consistent with serotonin’s absorption spectrum).22 However, a distinct class was found to repair its CPD with no dependence on serotonin, most optimally (as defined by kobs/kuncat) in the 300−310 nm wavelength range.22 Up until this point, the tacit assumption based on the published literature was that only a protein or an electron-rich small-molecule cofactor could catalyze CPD repair. The representative cofactor-independent DNAzyme that we ob-
high sun exposure show fewer bipyrimidine steps entirely.6 Of the various species of pyrimidine photodimers, the cis-syn cyclobutane thymine dimer (CPD) is the most commonly occurring.7 The mechanism of CPD formation is a photochemically allowed 2 + 2 cycloaddition reaction, although these dimers may also form via indirect excitation of a triplet photosensitizer, including those found within cells.8 In their natural environment, CPD photolyases are responsible for recognizing a CPD flipped out from a DNA double helix and, upon irradiation, repairing the dimer.9 Pioneering work with small-molecule analogues of the photolyase cofactors10,11 was instrumental for important mechanistic insights: all photolyase enzymes, including those that repair CPDs, contain a flavin adenine dinucleotide (FAD) cofactor, which is maintained in a fully reduced and anionic state within the holoenzyme. In the photolyase catalytic cycle, this cofactor becomes excited and transfers an electron to the CPD. Upon reduction, the CPD rapidly reverts to its constituent pyrimidine monomers and returns its excess electron to the flavin, restoring the latter’s resting form. The catalytic cycle can then be repeated. The essential flavin cofactor, in its anionic form, has a relatively low extinction coefficient (∼2 × 10−3 M−1 cm−1 at 450 nm) compared with flavin in its oxidized form (∼10 × 10−3 M−1 cm−1 at 450 nm).12 Photolyases therefore typically contain a second, chemically variable, prosthetic cofactor that acts as an antenna, absorbing additional light and transferring its energy nonradiatively to the flavin (Figure 1a). Class I photolyases,
Figure 1. Overview of (a) a photolyase enzyme and (b) The UV1C DNAzyme with photolyase activity. The DNAzyme itself is shown in black, while its DNA substrate, TDP, is shown in red.
such as from Escherichia coli or yeast, use a folic acid derivative, MTFH, as the antenna and show maximal activity at ∼380 nm. Class II photolyases, such as from Aspergillus nidulans, use a deazaflavin antenna cofactor, with resulting maximal activity at ∼440 nm.13 In rarer instances, flavin mononucleotide (FMN) or even a second unreduced FAD serve as antenna cofactors. The presence of such antennae greatly enhances the photocatalytic cross section of a photolyase;14 nevertheless, when both flavin and antenna cofactors are removed, photolyases remain vestigially functional. This is attributable to a tryptophan residue near the CPD binding site.15 Even simple peptides containing tryptophan are able to show some photorepair activity.16 These observations that tryptophan is able to play a functional role in CPD photoreactivation led Schultz and colleagues17 to hypothesize that an antibody raised against CPD could position a tryptophan close to CPD bound in its antigen-binding site, enabling photoreactivation of the CPD. A series of monoclonal catalytic antibodies, or “abzymes”, raised from such an underlying premise displayed efficiency 527
DOI: 10.1021/acs.accounts.7b00582 Acc. Chem. Res. 2018, 51, 526−533
Article
Accounts of Chemical Research
Figure 2. (left) Rate enhancement (defined as kobs/kuncat) profiles of the DNAzymes of UV1C and Sero1C, shown as functions of irradiation wavelength. (right) Initial velocity of TDP photoreactivation by the UV1C DNAzyme plotted against the substrate (TDP) concentration.
Figure 3. Three paradigms of purely DNA-dependent thymine dimer repair within double helices, each invoking light-induced electron transfer. Electron transfer to an adjacent thymine dimer from (top) a photoexcited guanine, (middle) 8-oxoguanine, and (bottom) a guanine−adenine exciplex is shown.
serotonin-dependent DNAzyme Sero1C, are comparable at their distinct peak irradiation wavelengths (305 nm for UV1C and 315 nm for Sero1C). This observation may reflect upper limits for their catalytic capabilities imposed by the specific SELEX protocol used. Both UV1C and Sero1C are bona fide enzymes, turning over substrate multiple times.27 The right panel of Figure 2 shows a plot of reaction velocity versus substrate concentration for UV1C. The two DNAzymes, however, show very distinct substrate repertoires. While UV1C strongly prefers the original TDP substrate, Sero1C photoreactivates a number of different substrates. UV1C strongly prefers deoxyribose sugars and repairs CPDs in the following order: dTdT ≫ dTdU ≈ dUdT > dUdU > rTdT > dTrT (where “d” stands for deoxyribose and “r” for ribose). All of the other pyrimidine dimer combinations tested were poor substrates, including those involving cytosine.22,23 Sero1C, by contrast, is proficient at repairing a significant range of substrates, including ribouridine dimers, ribothymidine dimers, and deoxyuridine dimers, as long as they are embedded within a larger DNA substrate.22 Both UV1C and Sero1C photoreactivate thymine dimers more efficiently than they do uracil dimers.21−23
tained (called UV1C) was subjected to our intensive study. On the basis of mutational, chemical foot printing and contactcross-linking experiments, it became clear that the structural and functional core of the UV1C DNAzyme was a two-layered G-quadruplex21,23 (Figure 1b). While the nonaromatic CPD dimer has effectively no absorbance at light wavelengths above 240 nm,2 UV1C’s action spectrum stretches fully to the red edge of DNA’s absorption spectrum (to ∼310 nm). UV1C’s photoreactivation quantum yield (Φ) measured at 305 nm was 0.05.21 While this is significantly lower than the Φ value of ∼0.7 found for CPD photolyases, it is close to those of 6−4 photolyases (Φ ∼ 0.1)24 and the catalytic antibody (Φ ∼ 0.08).18 It is also far higher than the quantum yield of the background reaction (Φ ∼ 10−4).21 The above observations cumulatively suggested that photoexcited electron transfer was the mechanism for UV1C’s photoreactivation of CPDs.21 Indeed, two distinct kinds of contact cross-linking showed the extreme proximity of the CPD within the bound TDP substrate to UV1C’s G-quadruplex.17,25 Such proximity would be necessary for efficient electron transfer from the G-quadruplex to the CPD.26 The left panel of Figure 2 shows that peak catalytic enhancement by the two DNAzymes, UV1C and the 528
DOI: 10.1021/acs.accounts.7b00582 Acc. Chem. Res. 2018, 51, 526−533
Article
Accounts of Chemical Research
electron−hole recombination, they exist in the Marcus inverted region and are inhibited from rapid decay to the neutral ground state.31 Examining CPD repair in short oligonucleotides, the authors found that the sequence GATT repaired itself 7 times faster than TT under 266 nm irradiation, while TAT T showed no enhanced self-repair at all (Figure 3 bottom). Such a pattern is explained by the requirement that for repair to occur, the excimer must form in a way that places the excess negative charge on the base 5′-adjacent to the dimer.32 If the charge-separated excited states were not in existence for such a relatively long time, the excess electron would have much less chance to jump to and repair the neutral dimer. Comparing the results of these reports to ours, we note that the local sequence around the CPD located within the UV1C substrate TDP is TATT, consistent with our observations of a low efficiency of uncatalyzed repair.21 The above set of reports raised several points of interest. First, if 8-OG is capable of shifting the action spectrum of the G-dependent repair mechanism within duplex DNA to wavelengths longer than 305 nm, can the action spectrum of the UV1C DNAzyme itself be tunable by selective incorporation of other chromophores into its sequence? Second, can UV1C, selected to photoreactivate a gapped CPD within its substrate, TDP, also repair a CPD thymine dimer within a continuous, ungapped DNA strand (“LDP”, a single-stranded DNA of the same length and sequence as TDP but incorporating a phosphodiester between the dimerized thymidines)? Third, is UV1C a bona fide photolyase, catalyzing only photoreactivation of the thymine dimer within TDP (and, perhaps LDP), or, does it also catalyze the de novo formation of a thymine dimer within LDP?
Though the analogy to evolution is necessarily fraught, a comparison of the properties of UV1C and Sero1C, which arose out of the same in vitro selection experiment, offers a perspective about what may be functionally gained when a nucleic acid enzyme co-opts an amino acid-like cofactor (Sero1C, requiring serotonin for activity) relative to one that does not utilize a cofactor (UV1C). As shown in the left panel of Figure 2, the two DNAzymes offer almost identical rate enhancements for CPD reactivation. However, Sero1C has a substantially broader substrate repertoire than UV1C; therefore, in this small-scale “evolution” experiment, the co-opted cofactor contributes versatility to its DNAzyme rather than a gain in reaction efficiency. The concept that one DNA molecule (UV1C) can repair a CPD located within another DNA molecule (TDP) by means of photoactivated electron transfer has had a broad impact on thinking about photoelectronic processes in nucleic acids generally, including the double-helical DNA of living cells. The first such report was by Rokita and co-workers,28 who investigated the photostationary states of CPD formation within several short DNA duplex sequence contexts. These workers observed that the initial rates of CPD formation were similar in all of the sequences they studied; however, in sequences with a 5′ guanine adjacent to the CPD, the rate of CPD repair was significantly higher, leading to an lower overall yield of CPD (Figure 3 top). These results were consistent with our own findings in the case of UV1C, where the propensity of one or more guanine bases within the G-quadruplex to serve in photolyase-like CPD repair was clear. While guanine is the most readily oxidized of the four DNA bases, even in its ground state, the resulting oxidation product, 8-oxoguanine (8-OG), is a base of even lower oxidation potential. 8-OG was investigated by Burrows and colleagues29 for its ability to carry out photoactivated CPD repair, and indeed, it was found that 8-OG significantly enhanced the rate of thymine dimer repair when it was placed in the opposite strand from the CPD and irradiated with >300 nm light (8OG’s absorption spectrum is red-shifted relative to that of guanine). The location and orientation of the 8-OG modulates its effect, with the highest repair rate being obtained from placement of the 8-OG on the same strand and 5′ to the CPD (Figure 3 top). A fascinating finding of this study is that when 8-OG is located in the strand complementary to the CPDcontaining substrate strand, in the presence of excess of substrate strand each 8-OG is able to repair an average of five CPDs, although such turnover requires heating and cooling for duplex melting and reannealing. Thus, while not enzymatic behavior in itself, the above indicates that 8-OG is both donating electrons to the CPD and accepting electrons back from the repaired thymine products. Recently, Carell, Zinth, and co-workers30 measured the charge transfer dynamics of small DNA oligonucleotides following irradiation by UV light. Using 5-methylcytosine (5mC), which like 8-OG has a significant absorbance at 300 nm (at this wavelength neither adenine nor uracil absorb significantly). It was found that excitation of a single-stranded oligonucleotide containing a single 5-mC base gave decay profiles that were consistent with the formation of radical cation and anion states of the adenine and uracil bases. The authors noted that such long-lived, charge-separated “excimer” states may cause “oxidative or reductive damage currently not considered in DNA photochemistry.” While the chargeseparated excited states have a high driving force for
■
TOWARD VISIBLE LIGHT Proteinaceous photolyases make efficient use of visible light to repair CPDs. However, both the UV1C and Sero1C DNAzymes are capable of using only
