One of the detrimental effects of UV radiation on DNA is the formation of the (6-4) photoproduct (6-4PP) between two adjacent pyrimidines1. This lesion interferes with replication and transcription and may result in mutation and cell death2. In many organisms a flavoenzyme called photolyase uses blue light energy to repair the 6-4PP3. The molecular mechanism of the repair reaction is poorly understood. Here, we use ultrafast spectroscopy to show that the key step in the repair photocycle is a cyclic proton transfer between the enzyme and the substrate. By femtosecond synchronization of the enzymatic dynamics with the repair function, we followed the function evolution and observed direct electron transfer from the excited flavin cofactor to the 6-4PP in 225 ps but surprisingly fast back electron transfer in 50 ps without repair. Strikingly, we found that the catalytic proton transfer between a histidine residue in the active site and the 6-4PP, induced by the initial photoinduced electron transfer from the excited flavin cofactor to 6-4PP, occurs in 425 ps and leads to 6-4PP repair in tens of nanoseconds. These key dynamics define the repair photocycle and explain the underlying molecular mechanism of the enzyme’s modest efficiency.

Photolyase uses blue light to restore the major ultraviolet (UV)-induced DNA damage, the cyclobutane pyrimidine dimer (CPD), to two normal bases by splitting the cyclobutane ring. Our earlier studies showed that the overall repair is completed in 700 ps through a cyclic electron-transfer radical mechanism. However, the two fundamental processes, electron-tunneling pathways and cyclobutane ring splitting, were not resolved. Here, we use ultrafast UV absorption spectroscopy to show that the CPD splits in two sequential steps within 90 ps and the electron tunnels between the cofactor and substrate through a remarkable route with an intervening adenine. Site-directed mutagenesis reveals that the active-site residues are critical to achieving high repair efficiency, a unique electrostatic environment to optimize the redox potentials and local flexibility, and thus balance all catalytic reactions to maximize enzyme activity. These key findings reveal the complete spatio-temporal molecular picture of CPD repair by photolyase and elucidate the underlying molecular mechanism of the enzyme's high repair efficiency.DNA repair photocycle | ultrafast enzyme dynamics | thymine dimer splitting | electron tunneling pathway | active-site mutation U ltraviolet (UV) component of sunlight irradiation causes DNA damage by inducing the formation of cyclobutane pyrimidine dimer (CPD), which is mutagenic and a leading cause of skin cancer (1-3). CPD can be completely restored by a photoenzyme, photolyase, through absorption of visible blue light (4). In our early work (5-7), we have observed a cyclic electron-transfer (ET) reaction in thymine dimer (ThiT) repair by photolyase and determined the time scale of 700 ps for the complete repair photocycle (7). However, the central questions of whether the splitting of the cyclobutane ring is synchronously or asynchronously concerted or stepwise and whether the cyclic ET involves specific tunneling pathways were not resolved. Furthermore, the molecular mechanism underlying the high repair efficiency has not been elucidated. Here, using femtosecond spectroscopy and site-directed mutagenesis, we are able to measure the dynamics of all initial reactants, reaction intermediates, and final products with different substrates and with wild-type and active-site mutant enzymes, and thus reveal the complete spatio-temporal molecular picture of thymine dimer repair by photolyase.Photolyase contains a fully reduced flavin adenine dinucleotide (FADH − ) as the catalytic cofactor and electron donor (4). Based on previous studies (4-10), a sequential repair mechanism of thymine dimer splitting is shown in Fig. 1. Previously, we found that the forward ET from FADH − Ã to ThiT occurs in 250 ps (1∕k FET ) and the total decay of intermediate FADH• in 700 ps (1∕k total ) (5, 7). These dynamics usually follow a stretched-exponential decay behavior, reflecting heterogeneous ET dynamics controlled by the active-site solvation (5, 6, 11). However, in that study, no thymine-related species could be detected in the vi...

Although unsupervised person re-identification (RE-ID) has drawn increasing research attentions due to its potential to address the scalability problem of supervised RE-ID models, it is very challenging to learn discriminative information in the absence of pairwise labels across disjoint camera views. To overcome this problem, we propose a deep model for the soft multilabel learning for unsupervised RE-ID. The idea is to learn a soft multilabel (real-valued label likelihood vector) for each unlabeled person by comparing (and representing) the unlabeled person with a set of known reference persons from an auxiliary domain. We propose the soft multilabel-guided hard negative mining to learn a discriminative embedding for the unlabeled target domain by exploring the similarity consistency of the visual features and the soft multilabels of unlabeled target pairs. Since most target pairs are cross-view pairs, we develop the cross-view consistent soft multilabel learning to achieve the learning goal that the soft multilabels are consistently good across different camera views. To enable effecient soft multilabel learning, we introduce the reference agent learning to represent each reference person by a reference agent in a joint embedding. We evaluate our unified deep model on Market-1501 and DukeMTMC-reID. Our model outperforms the state-of-the-art unsupervised RE-ID methods by clear margins. Code is available at https://github.com/KovenYu/MAR.

Most existing Re-IDentification (Re-ID) methods are highly dependent on precise bounding boxes that enable images to be aligned with each other. However, due to the challenging practical scenarios, current detection models often produce inaccurate bounding boxes, which inevitably degenerate the performance of existing Re-ID algorithms. In this paper, we propose a novel coarse-to-fine pyramid model to relax the need of bounding boxes, which not only incorporates local and global information, but also integrates the gradual cues between them. The pyramid model is able to match at different scales and then search for the correct image of the same identity, even when the image pairs are not aligned. In addition, in order to learn discriminative identity representation, we explore a dynamic training scheme to seamlessly unify two losses and extract appropriate shared information between them. Experimental results clearly demonstrate that the proposed method achieves the state-of-the-art results on three datasets. Especially, our approach exceeds the current best method by 9.5% on the most challenging CUHK03 dataset.

The flavin cofactor in photoenzyme photolyase and photoreceptor cryptochrome may exist in an oxidized state and should be converted into reduced state(s) for biological functions. Such redox changes can be efficiently achieved by photoinduced electron transfer (ET) through a series of aromatic residues in the enzyme. Here, we report our complete characterization of photoreduction dynamics of photolyase with femtosecond resolution. With various site-directed mutations, we identified all possible electron donors in the enzyme and determined their ET timescales. The excited cofactor behaves as an electron sink to draw electron flow from a series of encircling aromatic molecules in three distinct layers from the active site in the center to the protein surface. The dominant electron flow follows the conserved tryptophan triad in a hopping pathway across the layers with multiple tunneling steps. These ET dynamics occur ultrafast in less than 150 ps and are strongly coupled with local protein and solvent relaxations. The reverse electron flow from the flavin is slow and in the nanosecond range to ensure high reduction efficiency. With 12 experimentally determined elementary ET steps and 6 ET reaction pairs, the enzyme exhibits a distinct reduction-potential gradient along the same aromatic residues with favorable reorganization energies to drive a highly unidirectional electron flow toward the active-site center from the protein surface.protein electron transfer | flavin photoreduction | femtosecond dynamics | electron flow directionality | reduction potential funnel P hotolyase and cryptochrome are evolutionally related and contain a flavin adenine dinucleotide (FAD) as the catalytic cofactor with a unique bent structure in the active sites, but the two perform different functions: photolyase repairs UV-damaged DNA and cryptochrome functions as a photoreceptor for regulation of plant growth or synchronization of circadian rhythm (1-5). The active state of the cofactor in vivo is in the anionic hydroquinone form (FADH -) in photolyase (6), but currently the redox status of flavin in cryptochrome is under debate with some studies suggesting flavin to be in oxidized (FAD), whereas others claiming anionic (FAD -/FADH -) states for the functional form in vivo (7-10). However, in vitro, the cofactor is oxidized to FAD and/or FADH • in photolyase but only appears in the oxidized FAD state in cryptochrome. Thus, it appears that reduction of the oxidized FAD is necessary for its transformation into the active state. Photoinduced electron transfer (ET) is an effective way for such redox state changes and it is conceivable that the photoreduction of FAD could be a primary process for signal initiation in cryptochrome (11).In this study, we used Escherichia coli photolyase (EcPL) as a model system for systematic studies of electron flow into the excited cofactor FAD. As shown in Fig. 1, more than 10 aromatic acid residues (W and Y) encircle the flavin cofactor around the active site (12). These aromatic residues not only are cri...