Capturing Protein–Nucleic Acid Interactions by High‐Intensity Laser‐Induced Covalent Cross‐Linking†

Interactions of DNA with structural proteins such as histones, regulatory proteins and enzymes play a crucial role in major cellular processes such as transcription, replication and repair. The in vivo mapping and characterization of the binding sites of the involved biomolecules are of primary importance for a better understanding of genomic deployment that is implicated in tissue and developmental stage‐specific gene expression regulation. The most powerful and commonly used approach to date is immunoprecipitation of chemically cross‐linked chromatin (XChIP) coupled with sequencing analysis (ChIP‐seq). While the resolution and the sensitivity of the high‐throughput sequencing techniques have been constantly improved, little progress has been achieved in the cross‐linking step. Because of its low efficiency, the use of the conventional UVC lamps remains very limited while the formaldehyde method was established as the “gold standard” cross‐linking agent. Efficient biphotonic cross‐linking of directly interacting nucleic acid–protein complexes by a single short UV laser pulse has been introduced as an innovative technique for overcoming limitations of conventionally used chemical and photochemical approaches. In this survey, the main available methods including the laser approach are critically reviewed for their ability to generate DNA–protein cross‐links in vitro model systems and cells.


INTRODUCTION
Genomic DNA is constantly associated with various proteins, thus forming chromatin (1). In eukaryotic organisms, DNA is wrapped around octamers of core histones (two of each H2A, H2B, H3 and H4) to form the nucleosome. Upon binding of the linker histone H1 and various "architectural" proteins, the nucleosome filaments, consisting of connected nucleosomes through the linker DNA, fold up into higher order compact chromatin structures (1)(2)(3)(4). In addition to these structural proteins, DNA is associated with different functional proteins that are involved in essential cellular processes such as replication, transcription, repair or recombination. The expression of genetic information is a function of the reversible interactions of these proteins. A detailed mapping of protein-DNA interactions is needed for characterizing the molecular mechanisms involved in gene regulation. The majority of these interactions are transient and highly dynamic, which constitutes a major difficulty to preserve their integrity when subjected to biochemical purification and fractionation methods. This could be overcome by "fixing" the interactions via protein-DNA-induced covalent cross-linking (5,6).
Currently, induction of DNA-proteins cross-links (DPCs) is mainly obtained by conventional chemical methods. These involve the use of bifunctional reagents to generate DNA-protein crosslinking at the contact sites and "freezing" the interaction existing in the living cells (5). The formation of DPCs associated with chromatin immunoprecipitation (ChIP) allows further immunoselection of the DNA fragments that interact with the protein of interest (7,8). Coupling of chromatin immunoprecipitation with highthroughput techniques such as next-generation sequencing ChIPseq (9)(10)(11) led to substantial improvement of the analytical approach through the increase in the resolution and sensitivity thus extending the analysis of transcriptional regulation, epigenetics and chromatin structure into the genome-wide scale. This made possible the whole genome to be assayed in a single experiment with the generation of high signal-to-noise accessible digital data.
Further improvement of the resolution to the base-pair level was achieved by exonuclease-mediated digestion of DNA flanking the protein cross-linked site to create narrow footprint in its vicinity allowed by ChIP-exo (12) and ChIP-nexus (13). Formaldehyde (FA) is frequently employed as "the gold standard" cross-linking agent in ChIP-seq experiments because its reaction yields are typically larger than those provide by other methods. However, the in vivo formaldehyde fixation procedure remains still poorly understood, and little is known about the specificity of cross-linking, its efficiency and the structure of the resulting chemical adducts with the exception of a few of them that were identified. Several drawbacks related to the fact that formaldehyde is more efficient to produce protein-proteins rather than protein-DNA cross-links have been identified (see below). Thus, this makes it difficult to assess whether the observed protein association with DNA is a direct interaction with DNA or an indirect process involving other DNAbinding proteins, especially because of long incubation times required for formaldehyde cross-linking. Protein-protein crosslinking may lead to detection of interaction sites as artifacts, in particular false positive peaks at highly accessible loci (14,15). Excessive FA treatment can result in epitope masking (16) and preventing chromatin fragmentation (17) leading to native interaction peaks loss (false negatives). Thus, use of FA in cross-linking ChIP experiments can prevent accurate data interpretation (14,(18)(19)(20). UV light emitted by low-intensity germicidal lamps that produces "zero-length" covalent bond between contact points of nucleic acid and protein has proved to overcome some of the above-discussed limitations of formaldehyde-mediated crosslinking. The ChIP technique based on the use of low-intensity far ultraviolet (254 nm) irradiation for the generation of DPCs was introduced by Gilmour and Lis in the 1980s for mapping direct protein-DNA interactions in vivo (21,22). The sequencing method was also applied to study the interaction of several transcription factors with proteins at individual loci in Drosophila cells (23)(24)(25). Unfortunately, the low yields of cross-linked material generated in mammalian cells despite long irradiation times constitute a major limitation to the application of UV ChIP (26,27). It has been shown that thymine exhibits a higher ability to cross-link to proteins than other nucleobases (28,29), and thus, achievement of an efficient cross-linking reaction requires to check whether the binding site containing thymine residues is in a favorable orientation; this necessitates a validation step on model in vitro studies before performing in vivo experiments (30). Of note, both formaldehyde and UV lamp cross-linking methods suffer from the fact that the time required for the treatments are rather long (from several minutes up to several tens of minutes), precluding their use to study rapid kinetics. In addition, they may also create covalent links among structures that only casually come into ephemeral contact.
High-intensity UV laser irradiation represents an alternative to overcome these limitations. Indeed, delivery of photons in ultrashort nano-or picosecond time intervals has been shown to efficiently generate DPCs within a single ultrashort pulse either in vitro or within nuclei and whole cells (6,31,32). The use of a single ultrashort UV laser pulse instead of either a cross-linking chemical agent or low-intensity UVC lamp irradiation has a double advantage: First, it allows rapid "freezing" of the macromolecular interactions by applying "snapshots" at any step during the assembly of large protein-DNA complexes. Second, the excitation rate can be faster than the relaxation of the excited DNA bases thus allowing a second photon to be absorbed, giving rise through ionization to highly reactive nucleobase radical cations. These radical cations, in a subsequent step, are efficiently implicated in covalent adduct formation with amino acid side chains of specifically interacting proteins (for reviews see (33)(34)(35)). Initial attempts to determine whether high-intensity nanosecond laser UVC pulses could increase the efficiency of protein-nucleic acid cross-link formation through a biphotonic mechanism ended-up with contradictory results. Neither Harrison et al. (31) nor Hockensmith et al. (32) found an appreciable nanosecond laser intensity-dependent enhancement of DNA-protein cross-linking induction on model systems, thus ruling out involvement of a biphotonic photoprocess. However, later on, several other studies reported intensity-dependent quantum yields of cross-link formation. Using 266 nm nanosecond Nd:YAG laser pulses, Budowsky et al. measured a linear pulse dose response of the protein-RNA crosslinking quantum yield in the 30S subunit of E. coli ribosomes (36). Irradiation of the same sample with a continuous-wave source yielded 20-100 times lower total cross-links, with a different distribution of protein-RNA cross-links. Angelov et al. observed more than an order of magnitude increase in the quantum yield of histone-DNA cross-linking in reconstituted nucleosomes and whole cells with either 30 ps Nd:YAG laser pulses at 10 7 -10 9 W cm À2 (6) or with 5 ns laser pulses in the range of 10 6 -5.10 7 W cm À2 (5-250 mJ cm À2 ) (37). Similarly, Angelov et al. found an about eight-fold increase in the quantum yield of NF-jB-p50-DNA cross-linking in the range of 2.10 6 -5.10 7 W cm À2 (10-250 mJ cm À2 ) with 5 ns Nd:YAG laser pulses (38). Dobrov et al. confirmed the occurrence of two-quantum (intensity-dependent quantum yield) photo-induced RNA-protein cross-linking in tobacco mosaic virus with either 10 ns Nd:YAG laser (10 6 W cm À2 ) or 23 ps Nd:YAG laser (1.510 9 W cm À2 ) but with a higher efficiency under the picosecond laser regime (39). Russmann et al. reported an increase in the quantum yield of protein-DNA cross-linking with the decrease in the laser pulse duration (40). Comparative experiments were performed using 3 different laser sources on purified progesterone receptor and an oligonucleotide containing its consensus DNA-binding sequence. Thus, total cross-linking yields of approximately 1%, 4% and 30% were measured for irradiations performed with 5 ns Nd:YAG laser, 100 ps and 200 fs Ti:Sapphire lasers, respectively. Interestingly, as the peak intensity of the 200 fs pulses was 1000 times lower than that of the other two lasers, the femtosecond laser data does not support a two-photon-based mechanism. In contrast, in a more recent study (41), no increase in the femtosecond laser crosslinking quantum yield was noted with respect to low-intensity lamp values, thereby denying any significant involvement of twoquantum chemistry.
Overall, these experiments provided highly contradictory data and controversial response on the benefit of using high-intensity lasers delivering short UV pulses instead of conventional UV lamp sources to enhance the protein-nucleic acids cross-linking yield. The purpose of the present review is to critically analyze the experimental details, considering some basic principles of the laser photonics to uncover the source of this controversial issue and to provide some useful recommendations that should be considered for properly designing laser ChIP experiments.
to generate DNA-amino acid monoadducts (44,45) and crosslinks between proteins and DNA (5,46,47). The latter property has been used for the detection of DNA-protein binding contacts (48) in conjunction with an immunoprecipitation step followed by deep sequencing called ChIP-seq (49,50). To perform ChIPseq, cells are treated with FA thus resulting in the covalent cross-linking of proteins to DNA sequences at sites where they are supposed to interact. The DNA is then fragmented and the protein-DNA complex is isolated by immunoprecipitation using an antibody directed against the protein of interest. FA can react with the exocyclic amino group of guanine, adenine and cytosine to generate a methylol adduct that partially dehydrates to produce a highly reactive Schiff base toward amino and sulfhydryl groups (51,52). Thus, nucleophilic addition of reactive amino acids including lysine, tryptophan, arginine and cysteine gives rise to DPCs through the formation of a methylene bridge in model studies and cells (51,53). Guanine-lysine adducts that were preferentially formed are however unstable. This is not the case for cysteine adducts to adenine, cytosine and guanine. In addition, histidine and tryptophan were found to cross-link to adenine and guanine respectively. As an alternative route for generating DPCs, FA is able to react with lysine and tryptophan. The resulting Schiff base may further react with the exocyclic amino group of nucleobases to produce protein adducts to DNA.
Formaldehyde cross-linking associated with chromatin immunoprecipitation (FA-ChIP) is nowadays the commonly used technique for mapping nuclear proteins in the genome. However, macromolecules can be potentially cross-linked together in multiple ways and combinations, forming locally heterogeneous larger daisy-chained high-order chromatin or nuclear structures that cannot be completely destroyed even by prolonger sonication in presence of SDS. Consequently, the sonicated and immunoprecipitated DNA fragments may be associated with indirectly interacting more or less proximal proteins through protein-protein cross-linking bridges, as occurring in 3C experiments, thus yielding misleading interpretation.
Importantly, as illustrated in Fig. 1, the yield of formaldehyde-generated protein-protein adducts greatly exceeds that of protein-DNA adducts. Sonicated cellular DNA fractions 8-12 sediment as single-protein-DNA cross-links generated by UVC laser irradiation (Fig. 1A) in contrast with the majority of formaldehyde multiprotein-DNA covalent complexes that sediment as the protein fractions close to the top of the CsCl gradient ( Fig. 1B) (54). These CsCl sedimentation profiles are preserved under both heavy (Fig. 1C) and insufficient (Fig. 1D) FA cross-linking (lower panels) together with the DNA gel electrophoresis profiles of the corresponding fractions (upper panels) (55). Note that even at low level (single-hit conditions) of crosslinking (Fig. 1D), DNA-protein adducts are found at the top of the gradient similar to the well cross-linked-heavily sonicated DNA fragments (Fig. 1C). This suggests that, even at a low level of DNA-protein cross-linking, no single proteins but covalently linked multiprotein complexes are most likely captured by the FA cell treatment. The large number of peaks of unrelated proteins, found by FA-ChIP-seq, especially in regulatory and transcribed regions of active genes is also highly suggestive of a technical issue that is inherent to this cross-linking method. Under usual FA-ChIP conditions, the majority of formaldehyde TFs peaks do not correspond to consensus binding sites (56) suggesting that they represent false positive signals. Interestingly, FA treatment is able to generate cross-links even for non-DNA-interacting proteins such as GFP in an expression-dependent manner (14,57). Induction of the GAL genes resulted in an increased ChIP signal of the GFP protein and entire silencing complexes at the active loci, with presumably no biological relevance. In addition, wide range of experiments can be biased by the potential for formaldehyde to affect antibody recognition and epitope loss as observed in (56) representing false negative cases. Besides, the long incubation time (ten to several tens of minutes) restricts FA cross-linking to stationary study and makes it incompatible with dynamic interaction capture of protein-DNA complexes. The latter may be due in part to the harsh sonication treatment used to destroy the formaldehyde-created network that contributes to cross-linking reversal leading to loss of direct contacts. Besides, the long incubation time (ten to several tens of minutes) restricts FA cross-linking to stationary study and makes it incompatible with dynamic interaction capture of protein-DNA complexes. Briefly, whereas ChIP is a broadly valuable technique, it is clear that the use of formaldehyde in cross-linking could cause major issues in the interpretation of the experimental data and conclusions in many published papers may deserve reevaluation (19).

UV CROSS-LINKING
Irradiation with 254 nm ultraviolet (UV) light leads to the formation of a covalent bond between proteins and nucleic acids within cells, thereby covalently "fixing" the interactions between the two biopolymers (58). UV light-induced cross-linking of proteins to nucleic acids has been demonstrated for a variety of nucleic acid-binding proteins (for a review see (28)). Many model systems employed to address the chemistry of this photoinduced process (59)(60)(61) have shown that the mechanistic pathways involved are complex with widely competitive reaction pathways and variable yields. These studies have also shown that any amino residue can, in principle, be covalently UV crosslinked with any nucleobase residue of DNA or RNA (for a review see (28)). Thus, despite the fact that reaction mechanisms are far from being well understood, the generation of covalent bonds between proteins and DNA by conventional 254 nm UV radiation represents a useful and reproducible technique that has been applied successfully in several in vitro and in vivo studies of protein-DNA interactions (21,22,62,63). However, due to the poor photochemical efficiency leading to very low DNA-protein cross-linking yields, this approach has found only limited applications, thus explaining why the formaldehyde-mediated crosslinking has become a popular method in the field.
Our primary motivation for developing the UV laserbiphotonic ionization version of the cross-linking methodology was its unique potential to rapidly and stably capture native protein interactions by a single laser pulse. This allows nonstationary kinetics investigation by means of time-delay sampling following a cellular event initiation. Providing that the equilibria could be frozen, we also hoped to use such procedures to characterize in vitro and in vivo sequential interactions that occur within protein-nucleic acid complexes of biological interest. Moreover, the use of a single pulse allows avoiding artifacts due to cross-linking of UV damaged molecules. This is especially important for in vivo studies, to prevent artifacts from long conventional UV irradiation or formaldehyde incubation time affecting the entire cell metabolism. Other advantages of biphotonic UV laser-induced DNA-protein cross-linking include important enhancement of the quantum efficiency of the photo-induced cross-linking reaction and a larger spectrum of DNA-protein adducts produced with respect to conventional UV radiation excited state chemistry (for review see (34)).

Photophysical considerations
The idea that multiple-quantum excitation/ionization of biological molecules, namely nucleic acids (64), can be produced by ultrashort laser pulses of low-energy but high-peak power was formulated in the mid-seventies (65). This was proposed soon after multiquantum processes had been observed in the gaseous phase (66). Nucleic acids and their monomeric compounds in solution are characterized by a very rapid nonradiative decay of singlet excited states S 1 (Fig. 2). To absorb a second photon from S 1 , the value of the excitation rate must be close to the inverse value of its relaxation rate 1/s 1~1 0 12 s À1 . Such a resonant twoquantum (two-step) excitation can, therefore, be realized only by means of ultrashort high-peak power laser pulses. The situation is less critical if the second photon is absorbed by the triplet state T 1 with a comparatively long lifetime (I0 À6 s). In this case, a two-step excitation can be achieved by means of nanosecond laser pulses, however, with a reduced efficiency due to the low intersystem crossing yield that has to be compensated by a respective increase in the energy density. Under such irradiation, the molecule can absorb two 266-nm ultraviolet photons with a total energy of~9 eV. This largely exceeds the ionization potential of all nucleobases thus leading to concomitant electron ejection and radical cation formation. Note that the direct onequantum excitation of nucleobases to a similar energy state at lower wavelength (k = 133 nm) is impossible because of the strong absorption of the solvent water and buffers components.
Let us consider the simplified general scheme of molecular photoexcitation by absorption of one or two photons ( Fig. 2A). At a low-intensity, such as provided by a CW lamp or laser, a molecule can absorb no more than one photon because the excitation rate W exc = r 1 I is much lower than the relaxation rate 1/ s rel of the excited molecule (r 1 is the absorption cross-section, I the radiation intensity in photons cm À2 s À1 , whereas s 1 and s T are excitation relaxation times of S 1 and T 1 respectively). The molecule can take part in a chemical reaction either by direct excitation of S 1 or T 1 or indirectly by energy transfer to another molecule (e.g. oxygen). This scheme constitutes the basis for the linear, one-photon photochemistry (photobiology). Its main characteristic, the photoreaction quantum yield (i.e. the number of photo-reacted molecules per absorbed photons), is constant versus the intensity which is a direct consequence of the Bunsen-Roscoe law of Reciprocity. However, under short high-intensity laser pulses, a molecule can absorb a second photon either from an excited singlet or triplet state. This second scheme belongs to the nonlinear bi-(multi-) photon photochemistry and photobiology. These two completely different reaction pathways that impose different requirements on the laser intensity can lead to completely different chemical outcomes. One-photon absorption involves excited state chemistry of nucleic acids leading to a number of photoproducts such as pyrimidine dimers and 6-4 pyrimidine-pyrimidone adducts, DPCs etc., which is now a classical field encompassing thousands of published papers and many excellent reviews and books. Hence, we will not consider it here in detail but focuses our attention on the nucleobase radical cations chemistry resulting from biphotonic mechanisms. Excited state absorption results in excitation of the S N or T N Rydberg states situated above ionization threshold E i , leading to an electron ejection in the bulk and to the formation of highly reactive base radical cation.
Single-base resolution sequencing gel analysis together with high performance-liquid chromatography (HPLC) measurements and time-resolved spectroscopy provided support for the involvement of a charge transfer phenomena along the DNA duplex leading to preferential hole trapping by proximal guanine residues ( Fig. 2B), which are the main targets of two-photon ionization of DNA (67,(72)(73)(74)(75). Hydration of the guanine radical cation (G •+ ) leads to the generation of 8-hydroxy-7,8-dihydroguanyl radical that is converted upon one-electron oxidation into 8-oxo-7,8-dihydroguanine (8-oxoG) (76-78) while competitive oneelectron reduction gives rise to 2,6-diamino-4-hydroxy-5formamidopyrimidine (FapyG) (79,80). In addition, deprotonation of G •+ generates an oxidizing guanyl radical (G(-H) • ) that preferentially reacts with superoxide anion radical to produce 2,2,4-triaminooxazolone through the intermediacy of an imidazolone derivative (81,82). It should be pointed out that nanoand picosecond 266 nm laser irradiation of other purine and pyrimidine 2 0 -deoxyribonucleosides in aerated aqueous solutions leads to the formation of several oxidized compounds through hydration and deprotonation of initially generated base radical cations (72,73,83). Interestingly, biphotonic ionization of the bases in DNA that is expected to occur with a similar efficiency leads to predominant formation of 8-oxoG at the expense of other base oxidation products (67,74,75). This was also observed in cellular DNA (84) and rationalized in terms of charge transfer with preferential redistribution of positive holes on guanine bases (85,86), as further discussed in the next section.
The so-called saturation parameter S, that is, the fraction of molecules in the intermediate excited state at the end of the laser pulse provides a good estimate for the probability of occurrence of excited state absorption thus determining the overall efficiency of the biphotonic process. Having in mind that the lifetime of the T 1 is longer than the pulse duration of the usual Q-switched or mode-locked lasers used, S has different expressions depending on the ratio s 1 /s p where s p is the laser pulse duration. For the sake of simplicity, an optically thin (low absorbing A < <1) irradiation layer is selected and the backward S 1 -S 0 induced transition is neglected. Note that unlike optically thick layers (A ≥ 1), under optically thin layer approximation (A < <1), the beam intensity does not vary within the irradiated pathlength thereby the number of photons absorbed per molecule is concentration independent. In other words, all molecules are exposed to the same beam intensity, which considerably simplifies mathematical modeling of bi-, multiphotonic processes.
100 ps time scale in (dT)18 and (dT)20 (87)(88)(89) has been observed by ultrafast pump-probe spectroscopy and tentatively assigned as a singlet np* state (excimer state in case of (dA)20 and a charge separation state between two thymine bases in case of (dT)20). iv Femtosecond laser and intermediate S 1 state s p~1 0 À13 s < s 1 : S fs = 1.3410 18 r 1 E = 30E.
The first successful experiments involving picosecond laser pulses to trigger two-quantum excitation were performed on aqueous solutions of nucleobases (90,91). Two-(multi-) quantum photochemical reactions in DNA have been observed in various biological systems such as monomeric compounds, viruses, yeasts and cells (for a review see (33) and (92)). This approach is currently applied to genetic and epigenetic studies, for example, in combination with micro irradiation of cells (93), for generating various types of oxidatively generated DNA damage (e.g. oxidized bases and cross-links (34)) allowing in vivo investigation of spatial organization, interaction mapping and functions of complex molecules such as proteins and nucleic acids.

One-electron oxidation DNA-protein cross-linking chemistry
Guanine, possessing the lowest ionization potential (94)(95)(96)(97), is the preferential DNA target for one-electron oxidation and most likely the main player in laser DNA-protein cross-linking. There are currently few data in the literature concerning the chemical structure of DNA-protein cross-links induced by oxidation reactions. DPCs can be formed by addition of suitable amino acids to the guanine radical cation, originating from one-electron oxidation of DNA. Indeed, amino acids containing nucleophilic groups can potentially covalently bind to the cation radical of guanine, in position C8 to form adducts. The nucleophilic groups which may be involved are in particular amines, such as lysine and arginine, and alcohols (in the case of serine).
The possibility for low-intensity UVC/UVB radiation to mediate mono-photoionization of purine bases, essentially guanine, has been reported (110)(111)(112). However, as shown from the measurement of final degradation products including 8-oxoG arising from the conversion of initially photo-generated guanine radical cation in isolated cellular DNA, the ionization process constitutes an ultra-minor contribution (less than 1%) to the overall induced DNA damage by UVC or UVB radiation (113)(114)(115). Therefore, low-intensity UVC/UVB radiation appears to be unsuitable for the formation of oxidatively generated DPCs. In addition, evidence for occurrence of charge transfer from either a distant generated guanine radical cation (G •+ ) or globally induced pyrimidine and purine radical cations has been provided experimentally and confirmed by computational studies (67,116). The redistribution of initially produced base radical cations along DNA strands was explained by preferential trapping of the positive hole at guanine sites that act as efficient sinks. The charge transport (CT) chemistry that operates through both localized and delocalized hopping mechanisms (117)(118)(119)(120)(121) has been shown to be modulated by the presence of modified bases (122)(123)(124) or bound proteins (125)(126)(127). In the latter case, tyrosine and tryptophan when participating as amino acids in binding contacts were found to be oxidized by charge migration from DNA (125,128).
Actually, it is commonly accepted that G •+ which is in dynamic equilibrium with its deprotonated form, the highly oxidizing guanine radical G(-H) • , is the main precursor of the major DNA photodamage induced by type I photosensitization and direct biphotonic ionization (75,83). As an extension of the efficient hydration reaction of G •+ that was first identified in doublestranded DNA (76), evidence for the occurrence of other nucleophilic reactions is now available on the basis of type I photosensitization and biphotonic UVC experiments (for reviews see (75,129)). These include formation of intrastrand (130,131) and interstrand cross-linkages (129) together with DNA-protein cross-links (DPCs) (132). Several examples of DPCs formation via initial formation of G •+ are reported through the use of the suitable flash quenching technique that allows one-electron oxidation of distant guanines (133) without providing further information on the structure of the amino acid adducts to guanine (134)(135)(136). Unambiguous characterization of the main stable addition product of tri-lysine oligopeptide (KKK) to the TGT trinucleotide upon one-electron oxidation of guanine by excited riboflavin in aerated aqueous solution was achieved (132). This has involved HPLC separation of the main DPC before its enzymatic digestion into 2 0 -deoxyribonucleaoside and assignment of the released 2 0 -deoxyguanosine-lysine adduct by 1 HNMR and electrospray ionization mass spectrometry (EIMS)-based exact measurements (132). The spectroscopic data thus gained agrees with the formation of a N Ɛ -(guanin-8-yl)-lysine cross-link involving the central lysine of KKK through its free Ɛ amino group (Fig. 3). Further mechanistic insights were gained from a detailed ab initio molecular dynamics simulation study involving protonated methylamine as the model amino acid (137). The cross-link formation is rationalized in terms of initial deprotonation of G •+ followed by hydrogen transfer from the ammonium -NH 3 + to G(-H) • , thus allowing its chemical repair and generation of a nitrogen-centered radical that binds to C8 of guanine. Similar cross-links were generated by the covalent attachment of three polyamines including putrescine, spermine and spermidine to DNA at the C8 of G •+ produced by riboflavinmediated photosensitization (138,139). It was also shown that beside guanines other bases moieties can be also eventually involved in the formation of DNA-amino acid adducts. Thus, cytosine was found to be involved in the covalent attachment to NF-jB (38). This may be the result of initial or rearranged formation of cytosine radical cation that is susceptible to nucleophilic addition as evidenced in model studies. Another possibility would involve hole-migration from ionized DNA to bound peptide as shown in (128). Thus, time-resolved flash quench experiments on oligonucleotide duplexes with the [Ru (phen)(bpy')(dppz)] 2+ type I photosensitizer tethered at one end have demonstrated the mobility of the electron hole from base radical cation to the Tyr residue of bound Lys-Tyr-Lys peptide. Covalent adduct formation between either thymine or adenine sites and the tyrosine of Lys-Tyr-Lys tripeptide was detected by mass spectrometry as a final irreversible product of this peptide-to-DNA electron-transfer chemistry. This experiment shows that not only DNA base radicals but also amino acid radicals can be precursors of covalent DNA-protein adducts.
Nucleic acid-protein cross-linking with nanosecond and picosecond laser pulses The first application of high-intensity UV laser irradiation (20 nanosecond 248 nm pulses) for the generation (single-pulse) of protein-DNA cross-linking was reported almost 40 years ago (31). For this purpose, a mixture of purified E. coli RNA polymerase and [ 3 H]-labeled T7 DNA, supposed to form nonspecific complexes, were irradiated. Surprisingly, contrary to earlier publications on high-intensity UV laser photolysis of nucleic acids and related monomeric compounds, no difference in the crosslinking efficiency was noted upon exposure to either highintensity nanosecond 248 nm pulses provided by a KrF excimer laser or the low-intensity CW radiation provided by a germicidal 254 nm low-pressure Hg lamp. This finding has led to the conclusion that laser-induced cross-linking resulted from a monophotonic photoreaction. However, this conclusion is questionable for the following reasons: (i) use of a quite different lamp and laser irradiation geometries together with a relatively low precision on the indirect filter binding assay for crosslinking quantification, (ii) use of inappropriate mechanistic study model system consisting of a mix of E. coli RNA polymerase and T7 DNA that forms transient nonspecific and undefined complexes, (iii) the laser pulse dose response of the quantum efficiency reported is lying in the saturation range 0.3-1.2 J cm À2 of the biphotonic quantum yield making its shape indistinguishable from that expected for a monophotonic process (67).
Nanosecond UVC laser pulses from a Nd:YAG laser have been used to induce DNA-protein cross-linking in a mixture of phage T4-coded gene 32 protein (single-stranded DNA-binding protein) with one of the four single-stranded 10-mer deoxyhomopolymers that form low-affinity unspecific complexes (32,140). Irradiation was performed on optical thin samples in Eppendorf tubes, and the cross-linking analysis was achieved by SDS-PAGE. A maximum~3% yield of cross-links was obtained with (dT) 10 at a total irradiation energy of up to~100 mJ, while the yields of (dC) 10 , (dA) 10 and (dG) 10 were about 2 orders of magnitude smaller. The authors suggested that the process is monophotonic, that is, identical with that implicated under conventional low-intensity UV lamp radiation. However, these conclusions should be considered with a great care. First, because precise defined, experimental conditions were not provided for each series of experiments and improper dosimetry terminology was used in the data presentation. For example, the main conclusion on the photoreaction mechanisms was based on the total irradiation energy response of the photo-cross-linking yield instead on the laser pulse dose response of the relative quantum yield, that is, the total yield versus the laser pulse dose, at constant total irradiation energy. Second, only quantified values are shown without the respective row gel images that would be required to evaluate the quality of the experiments. Third, it is possible that the used substrates are not appropriate for biphotonic cross-linking to occur.
The group of Budowsky compared the RNA-protein crosslinking induced by low-intensity continuous 254-nm irradiation and high-intensity 10 ns 266 nm laser pulses in isolated 30S subunit of E. coli MRE 600 ribosomes (36). Cross-linking of proteins to the [ 32 P]-labeled 16S RNA was quantified by the filter binding assay while the protein subunits were identified by 2dimensional SDS-PAGE. The transition from low intensity to nanosecond laser UV irradiation was accompanied by an important increase in cross-linking efficiency and a larger number of proteins cross-linked to the 16S RNA. It is worth noting the overwhelming involvement of guanine residues in laser crosslinking. This contrasts with the lowest reactivity of guanine base when cross-linking is monophotonically produced by a CW UV lamp. These results demonstrate that cross-linking generated by high-intensity UV nanosecond laser irradiation proceeds via biphotonic excitation mechanisms that are different from those involved in conventional low excited state chemistry.
Having shown that the high-intensity laser radiation lead to more cross-linking compared with conventional UV lamp the question has arisen, which type of laser system nanosecond or picosecond is more convenient. The efficiency of both irradiation regimes was assessed by a combination of nitrocellulose filter binding and analytical centrifugation assays and then compared with low-intensity UV irradiation of [ 3 H]-uridine labeled tobacco mosaic virus (TMV) RNA in situ (39). A multipulse regime of irradiation was applied on optical thin layers within a wide range of laser beam intensity. The estimations showed that on the transition from low-intensity lamp to nanosecond irradiation the quantum yield of RNA-protein cross-links increases~3 times at E = 0.05 J cm À2 , while under picosecond irradiation regime, at the same laser pulse dose, the increase was~15 times. Such increases in the quantum yields suggest a two-quantum nature in the majority of RNA-protein UV laser cross-linking photoreactions under both laser regimes of irradiations. At E = 0.05 J cm À2 , the corresponding values of the saturation parameter are S ns = 1.510 À2 ; S ps = 6.510 À2 respectively which fit well with the reported results. It is noteworthy that the estimated population of the intermediated S 1 state population (picosecond laser irradiation) is 4-5 times higher than that of T 1 (nanosecond laser irradiation) under the same energy fluence of 0.05 J cm À2 . This is close to that noted for the saturation fluence concerning the formation of 8-oxoG, the major DNA oneelectron oxidation product, for the two laser regimes, see Fig. 3 in (72).
The first laser chromatin immunoprecipitation (LChIP) experiments were performed on chicken erythrocytes or Ehrlich ascites carcinoma cells and isolated nuclei submitted to multipulse picosecond laser irradiation in optically thick layers at a total irradiation energy of about 5-10 absorbed photons per nucleotide (6). After chromatin sonication and hydroxyapatite column filtration, the cross-linked complexes were subsequently isolated from free DNA by CsCl gradient centrifugation (Fig. 4A). The histone content in the cross-linked fractions was probed by immuno-dot assay (Fig. 4B,C) using highly specific anti-histone antibodies whose responses have been calibrated. The finding that histone tail acetylation did not interfere with cross-linking to linker DNA has been exploited to address relevant physiological issues concerning the presence of the core and linker histones in actively transcribed or silenced enhancer-promoter regions of ribosomal genes (54). A relevant example focused on the chromatin structure study of ribosomal genes of X. laevis embryos, in which the ribosomal genes were actively transcribed. Thus, it was found on the basis of UV laser ChIP and Southern blot analysis of specific region that coding sequences, spacer enhancers and promoters were associated with core histones both in actively transcribed embryonic genes and their silent counterparts of erythrocytes. In contrast, the level of the linker histone H1 in active regions was significantly reduced (54).
In a series of experiments, high-intensity pico-and nanosecond laser pulses at 266 nm were used for in vitro studies of histone-DNA interactions in whole cells, isolated total chromatin and salt-dialysis reconstituted nucleosomes. Exposure of the samples as optical thick layers to picosecond UVC laser pulses (reviewed in (34)) and subsequent analysis by CsCl gradient centrifugation revealed cross-linking formation estimated at 15% of bound histones to DNA. The laser pulse intensity response of the cross-linking yield within the 510 7 -10 9 W cm À2 (E = 1.510 À3 -310 À2 J cm À2 ) range at a constant total irradiation dose (picosecond laser irradiation regime) increased >20 times in a sigmoidal curve shape demonstrating the involvement of a two-quantum process in the photoaddition reaction (6). It may be pointed out that the latter laser irradiation conditions did not lead to detectable formation of biomolecule damage including DNA double-strand breaks, histone degradation and histone-histone cross-links. Strong evidence was provided in favor of a biphotonic origin of cross-linking formation under single-pulse nanosecond laser irradiation in optically thin samples of reconstituted nucleosomes (37,141). Interestingly, the majority of crosslinks involves the linker DNA and the nonstructured histone tails, mainly H3, independently on their acetylation status. These peculiarities of histone-DNA cross-linking observed on in vitro reconstituted systems are preserved in nuclei and whole cells for both nano-and picosecond laser irradiation conditions (142).
Single-pulse nanosecond laser-mediated cross-linking in optically thin samples was used for studying the solution contact points of human necrosis factor NF-jB p50 homodimer bound to a 37 base pair DNA containing its pseudo-symmetrical MHC H-2 binding site (38). SDS-PAGE analysis (Fig. 5A) was performed, and the total yield (Fig. 5B) and quantum efficiency (Fig. 5C) were plotted as a function of the laser fluence (laser pulse energy density). The sigmoidal shape of the curve and the almost one order of magnitude enhancement of the high-intensity quantum efficiency provided unambiguous evidence for the biphotonic mechanism of nanosecond laser cross-linking. Sequencing PAGE contact points mapping of SDS gel purified cross-linked fraction submitted to proteinase K and 3 0 -5 0 exonuclease treatment confirmed the existing crystal structure contacts and revealed new contacts, located symmetrically 2-3 bases apart from the extremity of the recognition sequence on the two opposite DNA strands (Fig. 5D).
The first application of the combination of nanosecond UV laser-induced cross-linking and mass spectrometry to study interactions between DNA and protein was done on the complex of the single-stranded DNA-binding domain of rat DNA polymerase beta (pol beta-ss) with oligonucleotide d(ATATATA) (145). Cross-linking was induced by irradiation of the DNA-protein equimolar mix in a optically thick layer configuration with a single 25 mJ laser pulse. MALDI-TOF mass spectrometry was used as an analytical tool to identify HPLC-purified cross-linked peptides from proteolytic digests. The identified cross-linked peptides showing the presence of amino acids that are in contact with DNA fragment were in accordance with existing NMR structural data. More peptides in contact in several regions were identified, as compared to low-intensity UV cross-linking, showing the superiority of the laser cross-linking approach. The relatively low yields of cross-linked peptides reported can be explained by the inappropriate conditions of single-pulse irradiation of highly concentrated optical thick samples in Eppendorf tubes, where not the entire solution but only a little surface proximal fraction of the solution is effectively irradiated. In case where large amounts of cross-linked complexes are needed for the analyses, either a multipulse irradiation of optically thick samples in a cuvette under vigorous stirring (6), or a single-pulse irradiation of optically thin layer in a flow-through cell (146) is more appropriate.
High-intensity single-pulse nanosecond UV laser cross-linking was employed to study the real-time dynamic interaction of factors with in vitro reconstituted MMTV chromatin templates (147). It was observed through the measurement of the extent of cross- linking by immuno-dot assay that interplay of the glucocorticoid receptor (GR) and the ATP dependent chromatin remodeler hSWI/ SNF complex is both transient and periodic. Furthermore, GR transient binding is correlated with an internal reorganization of histones H2A and H2B that proceeds through a structural transition within the core without changes in H3-H4 tetramer conformation. Interestingly, the described in vitro periodic transition closely approximates the in vivo characterized interaction events. These experiments demonstrate the utility of the laser cross-linking technique for studying transient dynamic events. Detailed experimental protocol used in this work was also provided in a separate paper (148). However, the major limitation that was not mentioned in the provided protocol was the strong screening effect on the laser beam transmission throughout the solution by the presence of 1 mM ATP (optical density = 15), which absorbs more than 90% of the incident photons flux, thereby strongly reducing the crosslinked fraction. Besides, the addition of 25 mM c-S-ATP in some experiments for quickly stopping hSWI/SNF ATPase activity induces a 100% screening effect, thus preventing any cross-linking photoreaction to occur.
A preparative procedure for single-pulse nanosecond UV laser cross-linking of chromosomal and regulatory proteins to DNA within whole cells was developed (146). A specially designed laminar flow cell device interfaced with a nanosecond Nd:YAG laser was used to irradiate the necessary total volume (amount) of cell suspension by fractions, each one of them being exposed to a single pulse under optically thin layers conditions. This innovative approach represented an important technical advance enabling the treatment of large amounts of substrates while preserving the benefits of the single-pulse (short irradiation time etc.) regime. This was used in combination with chromatin affinity-precipitation techniques (ChAP-qPCR) to investigate the interaction of the androgen receptor AR with the regulatory elements of several AR-dependent genes in cell cycle synchronized, nontumorigenic, immortalized prostate cells line WPE-1-AR tag that constitutively express an exogenous AR receptor (149). Although the level of AR expression was constant during the cell cycle and its translocation unaffected by the stimulator R1881, the expression of AR-target genes in response to R1881 showed a cell cycle dependency that peaked at the G1 phase. Two different cross-linking techniques, namely formaldehyde and UV laser DNA-protein cross-linking, that allow distinguishing between direct and indirect bindings to chromatin DNA, were used to address whether AR productively binds to target gene regulatory sequences. Chromatin affinity-precipitation (ChAP-qPCR) data showed that the direct binding (LChAP) of AR to its cognate sequences occurs mostly during the G1 phase, whereas indirect binding (FA-ChAP) to the same sites could also take place during the S phase. Interestingly, the binding at G1 correlated with the direct binding of polymerase II to target genes. This occurs exclusively during this phase, thus suggesting that G1 expression of AR-target genes resulted from AR-induced transcriptional activation through a direct and productive interaction with the DNA of the control regions of these genes. At the same time, analysis and comparison of the UV laser and formaldehyde cross-linking data showed the presence of NCoR protein bound through indirect protein-protein interaction during the S phase to AR-target genes. Laser cross-linking played a decisive role for identifying NCoR, which is strongly reduced in G1 and interact in S phase with AR and chromatin at AR-target genes, as an AR-dependent gene expression inhibitor during the cell cycle. Note that these features of AR-dependent genes could not be revealed by only Figure 5. High-intensity nanosecond UV laser irradiation-mediated formation of efficient cross-linking of NF-jB to DNA via a biphotonic mechanism. A: Representative 12% SDS-PAGE of a single-pulse-irradiated NF-jB-DNA complex at increasing laser pulse fluence. B: NF-jB cross-linking yield (the percentage of protein cross-linked to DNA) dependency on laser pulse fluence. C: Quantum efficiency (QE) under linear absorption approximation of NF-jB-DNA cross-linking formation versus laser pulse fluence. The experimental points were fitted to a dependence reflecting a two-quantum process in optical thin sample (72). D: The primary sequence of the 37 bp oligonucleotide used in the binding reactions. The NF-jB recognition sequence À5 to +5 is also indicated. The arrows indicate the points of contact between Parts I and II of the NF-jB p50 recognition loop and DNA as suggested by UV laser cross-linking, whereas black dots correspond to base-specific contacts observed in the crystal structures of DNA-NF-jB p50 and p52 (143,144). Reprinted from (38)  considering formaldehyde ChIP data, thus requiring the use of laser cross-linking information.
RNA polymerase V (Pol V) long noncoding RNAs (lncRNAs) were proposed to guide ARGONAUTE4 (AGO4) to chromatin in the RNA-directed DNA methylation (RdDM) process of plants. Application of FA-ChIP alone was unable to provide evidence for involvement of direct AGO4-DNA interactions since such complexes could be the result of indirect cross-linking through the interaction with other proteins. In contrast, application of a single-pulse LChIP on nuclei isolated from plant cells provided direct evidence for functionally relevant Pol V mediated specific AGO4-DNA interaction at RdDM loci (150). This result suggested a model with a possible role of DNA in specifying AGO4-dependent DNA methylation, that could account for the strand-biased nature and high degree of specificity of DNA methylation in RdDM. One could imagine that the pool of AGO4-siRNA effector complexes associated with the AGO hook platforms would directly inspect both DNA strands for complementary base pairing as Pol V proceeds through transcription elongation. This result illustrates the unique ability of the LChIP technique for in vivo mapping of direct DNA-protein interactions that cannot be achieved by using FA-ChIP.
Recently, UV laser ChIP-seq has been successfully used by the group of Saluz for the genome-wide mapping of the sequence-specific transcription repressor BCL6 binding site distribution in a human diffuse large B-cell lymphoma (DLBCL) cell line OCI-Ly1 (56), information that was compared with available published FA-ChIP-seq data. Fractions of 2.510 7 cell suspensions in a 200 lL volume were submitted to a multiple E p = 0.15 J cm À2 pulse radiation at 266 nm delivered by a nanosecond Nd:YAG laser to a total irradiation energy E tot~3 J, average~7 absorbed photons par nucleotide (irradiation in optically thick layer geometry and dosimetry conditions similar to our early experiments (6)) to cross-link direct "zero-length" protein-DNA interactions. It was shown that UV LChIP-seq produces sensitive and precise protein-DNA-binding profiles, encompassing about 7000 BCL6 binding peaks highly enriched (72.9%) with canonical BCL6 DNA sequence motifs displaying sharply clustered median (AE 18 bp) around the peak summit. Using this technique, numerous previously undetectable direct BCL6 binding sites were found, particularly within heterochromatin condensed inaccessible areas. Interestingly, UV LChIP-seq BCL6 binding sites showed large discrepancies, such as weak promoter binding and low overlap with DNase I hypersensitive sites, information that contrasts with previously published BCL6 binding sites captured by FA-ChIP-seq on the same cell line. Some previously FA-ChIp-seq BCL6 binding sites were also revealed by LChIP-seq. In addition, numerous previously FA-ChIp-seq undetectable direct BCL6 binding sites were found by LChip-seq particularly within heterochromatin condensed inaccessible areas. These discrepancies were attributed to whether those were direct or indirect binding contacts of BCL6 at these genomic loci. Although LChIP-seq approach is unable to address per se the functionality of the cross-linked DNA sequence motifs, identification of the directly interacting genome-wide binding sites, most of which could not be fixed with FA-ChIP, strongly suggests a repressive role of BCL6 in the investigated cell lines.
With few exceptions, it seems firmly established that, in general, both pico-(310 À11 s) and nano-(10 À8 s) second laser pulses with energy density~50-150 mJ cm À2 induce specific DNA-protein cross-linking through biphotonic one-election oxidation mechanism within in vitro and in whole cells conditions. Compared with conventional UV cross-linking, this approach is characterized by a significant enhancement of the cross-linking efficiency and decreased formation of monophotonic DNA lesions (67,73). The latter point is important since conventional UV-generated pyrimidine adducts that arrest the polymerase extension interfere with further DNA amplification.
An interesting feature of high-intensity picosecond laser radiation is its possibility to induce hydroxyl radical (•OH) by direct two-photon ionization/dissociation of water molecules (151). This highly reactive oxygen species induces additional DNA damage including single-strand breaks (33) that may reduce downstream PCR amplification and interfere with cross-linking data analysis. Although the picosecond laser irradiation is more efficient in terms of quantum efficiency for the generation of DNA-protein cross-links, this advantage is compensated by significantly larger pulse energies and average power of the nanosecond lasers together with their higher simplicity and reliability of utilization. It may be added that the significantly lower investment and maintenance cost of nanosecond lasers are other arguments in favor of their use.

Nucleic acid-protein cross-linking by femtosecond laser pulses
At first glance, femtosecond lasers generating pulses with duration comparable or lower than the lifetime of S 1 (<10 À12 s) seems to the most advantageous in term of quantum efficiency. Efforts have been made to optimize the conditions of UV laser cross-linking between the recombinant progesterone receptor (PD) and an oligonucleotide that contains a hormone-responsive element (HRE) by focusing on the role of laser pulses duration (40). Three types of lasers were used: (i) A nanosecond Nd: YAG laser delivering 266 nm pulses with s p = 5 ns, pulse energy 20-80 mJ at 10 Hz repetition rate; (ii) A picosecond Ti: Sapphire-oscillator-amplifier system, delivering 266 nm 100 ps duration pulses, with an energy of 5-7 mJ per pulse at 10 Hz, repetition rate; (iii) a quasi-CW mode-locked femtosecond Ti: Sapphire laser, delivering 200 fs duration pulses at 240-280 nm, with 1 nJ pulse energy at a 82 MHz repetition rate. Irradiation was carried out on optically thin (transparent) samples (less than 30% absorption) with a laser beam diameter d % 0.35 cm (s % 0.1 cm 2 ). With this in mind and applying the typical laser energies, the values of the saturation parameters for the different lasers used were as follows: (i) nanosecond laser: pulse energy 30 mJ, pulse dose E = 0.3 J cm À2 ; S ns = 0.09. (ii) picosecond laser: 100 ps, pulse energy 3 mJ, pulse dose E = 0.03 J cm À2 ; S ps = 0.3E % 10 À2 , S* ps = 0.9φ*; (iii) femtosecond laser: pulse energy 10 À9 J, pulse dose E = 10 À8 J cm À2 , S fs = 310 À7 . The main experimental result was that all three types of lasers are able to induce intensity-dependent cross-links by a biphotonic mechanism. However, the efficiency of the nanosecond laser was 5 times and~30 times lower than that of the picosecond and femtosecond lasers respectively. These estimations are quite contradictory as under the given irradiation parameters of the femtosecond laser (E = 10 À8 J cm À2 pulse dose) only a monophotonic process is expected to operate similar to lowintensity 254 nm irradiation.
The enhanced efficiency of DNA-protein cross-linking formation by femtosecond laser irradiation that was reported by (40) has led Lejnine et al. to determine optimal femtosecond laser irradiation conditions for the generation of DNA-histone crosslinking in nuclei isolated from human 293 cells (152). Amplified Ti:Sapphire 60 fs laser pulses at 266 nm (1 kHz repetition rate, pulse dose E~10 À4 J cm À2 , pulse energy 510 À6 J) were used in the irradiation of optically thick aliquots of 100 lL (1 mg mL À1 nuclei) to a total irradiation energy from 0.3 up to 9 J per sample (1 0 to 30 0 irradiation time) equivalent to a~4 to 120 absorbed photons per nucleotide. Linker H1 and core H3 histone were found to be cross-linked with overall cross-linking efficiency similar to that provided by a low-intensity UV lamp. This is strongly suggestive of the involvement of a single quantum photoreaction in the femtosecond laser regime. This conclusion confirming the data in (41) is not surprising taking into account the very low value of the saturation parameters for S 1 under the irradiation conditions used S fs~3 10 À3 , that is, insufficient for the induction of efficient two-quantum cross-linking. Here again, the use of such a complicated and expensive laser device was dispensable and the same results would have been obtained with conventional UVC radiation provided by either a low-cost mercury lamp at 254 nm or a 266 nm light emitting diode (LED).
Attempts were subsequently made to resolve this controversial issue. For this purpose, three different purified transcription factors TATA-binding proteins (TBP), a glucocorticoid receptor (GR) and a heat shock factor (HSF) bound specifically to respective short nucleic acid probes were submitted to either a lowintensity 254 nm lamp or a 150 fs 265 nm UV laser pulses with an intensity of up to 1 GW cm À2 and a pulse dose E = 1.510 À4 J cm À2 provided by an amplified Ti:Sapphire laser source (41). Surprisingly, application of even higher peak intensity fs pulses than in the previous study (40) did not improve the cross-linking efficiency compared with low-intensity CW 254 nm irradiation. The absence of enhancement is consistent with the relatively low value of the saturation parameter S fs = 30E = 4.510 À3 that is one order of magnitude lower than the value of S ns under the usual irradiation conditions. These experiments provide solid evidence that the very low pulse energy of the femtosecond lasers used in cross-linking experiments is insufficient to induce biphotonic cross-linking thereby disproving earlier reports (40,153) concerning significant advantages of fs lasers over lower pulse duration lasers or conventional UV lamp sources. It is worthy to note these conclusions apply to the energetics of the femtosecond laser used so far and in no way should be generalized to the UV laser cross-linking.
Recently, Reim and colleagues examined DNA-protein interactions with the help of the UV femtosecond laser pulse technology in conjunction with state-of-the-art nano-LC-MS analysis of DNA-cross-linked tryptic peptides (154). The method utilized a high-repetition rate low pulse energy femtosecond UV laser that can cross-link protein-DNA interactions. Irradiation was followed by digestion with nucleases and peptidases that cleave DNA and protein before MS identification of the HPLC purified peptide-DNA adducts. This was applied to reconstituted nucleosomes with recombinant histone octamers and to purified recombinant NF1 and TBP factors in complex with DNA, with aim of high-resolution mapping of protein-DNA contacts. Unfortunately, laser irradiations were carried out on a mix of different conditions including either optically thin or thick samples, with different volumes in Eppendorf tubes apparently without continuous stirring during irradiation, which is mandatory in multipulse experiments. This, together with the rather confusing and imprecise method description, precludes any strong conclusion in favor of the claimed biphotonic excitation mechanism. The rule for avoiding confusion and misinterpretation is to evaluate the corresponding saturation parameter S fs . Using the values provided in the paper: laser beam cross-section s = 510 À2 cm 2 and pulse energy 10 À7 -10 À8 J (pulse dose E = 210 À6 -210 À7 J cm À2 ) we find S fs = 30E = 610 À6 -610 À7 . These very low values of S fully exclude any involvement of biphotonic effects in the formation of cross-links. Therefore, the reported results of fliX-MS should be neither quantitatively nor qualitatively different from those that would have been obtained by using a conventional UV light source.
A train of 1 kHz repetition rate low-energy 200 fs duration laser pulses at 270 nm has been recently used (155) to study RNA-protein interactions by UV cross-linking both in vitro and in whole cells. Stationary, nonreal-time-resolved DNA-protein interactions experiments, consisting of cross-linking yield measurement versus total irradiation energy (presented in terms of irradiation time), were used to evaluate equilibrium binding constants and kinetic parameters such as association and dissociation rate constants both in vitro and in cells. This was achieved indirectly by means of mathematical fitting procedures with simplified models containing several variable parameters. Surprisingly, in some cases, the number of variable parameters was equal with the number of the significant experimental points. Interestingly, it was reported, in contradiction with the results in (41), that in vitro femtosecond laser cross-linking displayed a higher efficiency than conventional UV lamp, in contrast with the lower efficiency of RNA degradation (strand breaks). Noteworthy, the latter finding is in contradiction with all available data on laserinduced oxidatively generated DNA damage, in particular strand breakage (33). At the same time, in vivo comparative experiments did not show important qualitative and quantitative differences between UV lamp and femtosecond laser cross-linking that should be the case if biphotonic laser cross-linking was really involved. Finally, considering the laser irradiation parameters used in these experiments: pulse energy density 3.610 À6 and 9.310 À6 J cm À2 (repetition rate 1000 Hz, average power 1 mW and 2.6 mW, laser beam cross-section s = 0.28 cm 2 ), we find S fs = 30E = 1.110 À4 and 2.810 À4 respectively. These relatively low values of the saturation parameter S preclude the occurrence of biphotonic photo-processes, which makes the reported laser power dependencies surprising and raise the question of the mechanism involved.

CONCLUSION
The goal of this critical survey was not to oppose in an exclusive way the different approaches of DNA-protein cross-linking but to put each of them into the right place in accordance with the objective(s) to be achieved. While formaldehyde can be used for simultaneously capturing both directly interacting or proximal proteins to DNA (RNA), UV radiation is a purely "zero-length" cross-linker of proteins associated by direct contact with DNA bases. The disadvantage of the former is the impossibility to provide a clear characterization of the observed peaks while the disadvantage of the latter is the very low cross-linking yield in case of conventional UV light source. Besides, both approaches require long incubation (irradiation) times which not only makes them inappropriate for direct real-time dynamic studies but can be a source of bias. These drawbacks can be overcome by using high-intensity UV laser pulses as a tool for efficient generation of DPCs through biphotonic ionization pathway by "singlepulse" experiments in optically thin samples. Depending on the pulse duration, three types of lasers have been used so far including nano-, pico-and femtosecond lasers. However, the involvement of biphotonic ionization in the cross-linking reactions has been a matter of debate. Our critical analysis here shows that in several cases, the controversy can be sometimes due to improperly dedicated/designed experiments or the use of inappropriate laser type. This is particularly true for femtosecond laser cross-linking, the most widely UV sources used during the last years. Due to the weak energy, density per pulse under the usual irradiation conditions not only are these lasers unable to induce efficient biphotonic ionization, but also their quasi-CW operation regime requires continuous irradiation during relatively long times that makes them inappropriate for real-time dynamic experiments. The use of the 266 nm radiation provided by the fourth harmonic of nanosecond Nd:YAG lasers is undoubtedly the most appropriate solution in terms of efficiency, cost and simplicity. Although the utility of our early times multipulse LChIP approach (6,34) has been recently confirmed and extended by LChIP-seq experiments (56), the use of a singlepulse irradiation regime in optically thin samples (146) is recommended to avoid eventual artifacts as successfully demonstrated by LChIP-qPCR (149,150) or in designing real-time dynamic experiments.