RESEARCH ARTICLES◥

PHOTOCHEMISTRY

Chemiexcitation of melanin derivativesinduces DNA photoproducts long afterUV exposureSanjay Premi,1 Silvia Wallisch,1 Camila M. Mano,1,2 Adam B. Weiner,1*Antonella Bacchiocchi,3 Kazumasa Wakamatsu,4 Etelvino J. H. Bechara,2,5†Ruth Halaban,3,6 Thierry Douki,7† Douglas E. Brash1,6‡

Mutations in sunlight-induced melanoma arise from cyclobutane pyrimidine dimers(CPDs), DNA photoproducts that are typically created picoseconds after an ultraviolet(UV) photon is absorbed at thymine or cytosine. We found that in melanocytes, CPDsare generated for >3 hours after exposure to UVA, a major component of the radiationin sunlight and in tanning beds. These “dark CPDs” constitute the majority of CPDs andinclude the cytosine-containing CPDs that initiate UV-signature C→Tmutations. DarkCPDs arise when UV-induced reactive oxygen and nitrogen species combine to excite anelectron in fragments of the pigment melanin. This creates a quantum triplet state thathas the energy of a UV photon but induces CPDs by energy transfer to DNA in aradiation-independent manner. Melanin may thus be carcinogenic as well as protectiveagainst cancer. These findings also validate the long-standing suggestion that chemicallygenerated excited electronic states are relevant to mammalian biology.

Thepigmentmelaninprotects against sunlight-inducedburns,DNAdamage, and skincancer.These attributes of melanin are due to anunusually broad absorption spectrum andperhaps radical scavenging activity (1). Yet

paradoxes abound. First, melanin is not solelybeneficial:Blondesandredheads,whohaveahigherratio of yellow pheomelanin to brown eumelaninin their skin and hair, have a greater risk for mel-anoma than dark-haired individuals (by a factorof 2 to 4), and the pheomelanin/eumelanin ratioaccounts for some of this risk (2). UVA-irradiatedmice do not develop melanoma if they lack mel-anin (3), and Braf-mutant mice develop 10 timesas many spontaneous melanomas if they carrythe pheomelanin-associatedMc1re/e allele (4). UV-irradiated melanin, especially pheomelanin, trig-gers apoptosis and production of reactive oxygenspecies (ROS) and DNA strand breaks (5, 6). Mel-anin synthesis itself generates ROS, especiallysynthesis of pheomelanin (7).

Yet melanin’s ROS-generating properties donot explain its role in melanoma development,because most mutations in human melanomasdisplay the UV signature of C→T substitutions atsites of adjacent pyrimidines (8, 9). Such muta-tions arise from cyclobutane pyrimidine dimers(CPDs), which join adjacent pyrimidines to dis-tort the DNA helix (10, 11); genetic disorders ofCPD repair, such as xeroderma pigmentosum, ele-vate the risk for childhood melanoma by fourorders of magnitude. The most energetic com-ponent of sunlight, UVB (280 to 315 nm), createsthese DNA photoproducts almost instantaneously,so the only obvious effect of melanin is shielding.The lower-energy UVA (315 to 400 nm), whichconstitutes ~95% of the ultraviolet energy thatpenetrates the atmosphere, also induces mela-noma in mice and in humans who use tanningbeds, but it is inefficient at making the cytosine-containing CPDs that cause C→T mutations(12, 13). We provide evidence for an unusualbiochemical pathway that resolves these onco-logical discrepancies by making melanin an ac-tive participant in CPD formation.

Delayed cyclobutane pyrimidine dimerinduction in melanocytes

As a positive control for an unrelated experi-ment, we exposed murine fibroblasts and mela-nocytes to radiation from a UVA lamp, using anenzyme-linked immunosorbent assay (ELISA) tomeasure the number of CPDs over time. Induc-tion of CPDs by direct UV absorption is completein one picosecond (11). Consistent with this, wefound that for fibroblasts and for melanocytesderived from albino mice, the peak of CPD induc-

tion was reached immediately upon exposure,followed by slow nucleotide excision repair (Fig.1A and fig. S1A). (14). Unexpectedly, melanin-containingmurinemelanocytes instead continuedto generate CPD for at least 3 hours after UVAexposure, at which point generation was offsetby DNA repair (Fig. 1, B and C). The negativeresult with albino melanocytes implicated mela-nin in the process and ruled out the involvementof photosensitizing compounds in the growthme-dium. We confirmed the delayed production ofCPDs by using a comet assay for DNA strandbreaks after treating lysed cells with a nucleotideexcision repair enzyme to induce DNA breaks atCPD sites (fig. S1B). This result also indicatedthat delayed CPDs resided in nuclear DNA. De-layed CPDs were also visible by immunofluores-cence in normal melanocytes, whereas albinomelanocytes showed only repair (fig. S1, C andD). We conclude that pigmented melanocytesinduce “dark CPDs” after UV exposure ends.We next examined whether dark CPD induc-

tion was masked by concurrent repair and thusmight bemore extensive than it appeared. Knock-ing down excision repair of CPDs by means ofsmall interfering RNA (siRNA) against Xpa orXpc transcripts (corresponding to genesmutatedin xeroderma pigmentosum) revealed that darkCPDs constituted half of all CPDs (Fig. 1D). Thisfactor of 2 increase in DNA photoproducts is sub-stantial because cancer-predisposed individualsin xeroderma pigmentosum complementationgroup D are only 60% defective in CPD repair(15). For the higher-energy UVB, the majority ofCPDs in melanocytes were created in the dark,even without repair knockdown (fig. S1, E to H).The UVB process was also dependent on mela-nin, and thus it differs from the delayed produc-tion of CPDs that has been reported occasionallyin other cell types after UVC or UVB exposure(16–18).Human melanocytes also generated dark CPDs

after UVA or UVB exposure (Fig. 1E and fig. S1I),although there was interindividual variation inthe response, particularly for UVA. This likely re-flects genetic differences between the donors.For human melanocytes in which there was amodest rate of CPD decline after exposure butwhich showed no obvious dark CPDs, the darkCPDs were revealed by siRNA knockdown of XPAorXPC transcripts (Fig. 1F). It was not possible todetermine whether individual variation was dueto variation in melanin type because of privacyrestrictions for newborn foreskin tissue.To examine CPD induction in vivo, we used

transgenic mice in which melanocytes remain inthe epidermis throughout life because of expres-sion of the Kit ligand in keratinocytes (K14-Kitl),thus mimicking human skin. After exposure ofthese mice to UVA, the level of epidermal CPDsat the 2-hour time point was 3 times the levelimmediately after exposure (Fig. 1, G andH).Mostepidermal cells were keratinocytes, which receivepigment from the melanocytes; this finding sug-gests that melanin content, rather than synthesis,was the crucial requirement for dark CPD induc-tion. Furthermore, in K14-Kitl mice homozygous

RESEARCH

842 20 FEBRUARY 2015 • VOL 347 ISSUE 6224 sciencemag.org SCIENCE

1Department of Therapeutic Radiology, Yale University School ofMedicine, New Haven, CT 06520, USA. 2Departamento deBioquímica, Instituto de Química, Universidade de São Paulo,São Paulo 05513-970 SP, Brazil. 3Department of Dermatology,Yale University School of Medicine, New Haven, CT 06520, USA.4Department of Chemistry, Fujita Health University School ofHealth Sciences, Toyoake, Aichi 470-1192, Japan.5Departamento de Ciências Exatas e da Terra, UniversidadeFederal de São Paulo, Diadema, São Paulo 09972-270 SP,Brazil. 6Yale Comprehensive Cancer Center, Yale UniversitySchool of Medicine, New Haven, CT 06520, USA. 7INAC/LCIBUMR-E3 CEA-UJF/Commissariat à l’Energie Atomique (CEA),38054 Grenoble Cedex 9, France.*Present address: Pritzker School of Medicine, University ofChicago, Chicago, IL 60637, USA. †These authors contributedequally to this work. ‡Corresponding author. E-mail: douglas.brash@yale.edu

on February 23, 2021

http://science.sciencem

ag.org/D

ownloaded from

for the inactive Mc1re/e allele, in which melano-cytes synthesize red-yellow pheomelanin, bothinitial and dark CPDs were twice as frequent as

in black mice. This suggests that pheomelanin isboth a poorer shield against normal CPD forma-tion and a more potent dark CPD generator.

Cytosine-containing dark CPDs

We next used mass spectrometry to identify in-dividual CPD species inmurinemelanocytes. Thisanalysis revealed that thedelayedCPDs inmelanin-containing melanocytes included the cytosine-containing CPDs that generate UV-signature C→Tmutations (Fig. 2, A and B). Unexpectedly, rela-tive to directly inducedCPDs,we observed ahigher(TC + CT)/TT CPD ratio (by a factor of 4) in thedelayed CPDs, which is suggestive of unusualchemistry. UVA generates primarily TT CPDs,with TC and CT CPDs together contributing only10 to 30% of these three CPD types (13). The (TC +CT)/TT CPD ratio in melanin-containing cellsincreased from the UVA-like 0.37 at 0 hours (di-rectly induced CPDs) to 1.33 in the CPDs arisingduring the next 2 hours (P = 0.027)—a ratio moretypical of cells irradiated with the higher-energyUVB. Mass spectrometry ruled out two possibil-ities: (i) that the ELISA and comet assays cross-reacted with melanin-DNA adducts, and (ii) thatthe delayed CPD appearance reflected greateraccess of the antibody or endonuclease due to achange in DNA conformation after UV exposure.

Photochemical pathway

To identify the photochemical pathway that pro-duces dark CPDs, we focused on UVA because itreduces background fromCPDs created by directUVB absorption, reduces photosensitization fromaromatic molecules, and is used in indoor tanningbeds. We found that kojic acid—an inhibitor oftyrosinase, the rate-limiting enzyme in mela-nin synthesis—suppressed production of darkCPDs by 85% (Fig. 3A). Because ROS can beproduced in cells for long periods after UVAexposure (19), we tested the ROS scavengers

SCIENCE sciencemag.org 20 FEBRUARY 2015 • VOL 347 ISSUE 6224 843

Fig. 2. Mass spectrometryreveals enrichment of darkCPDs for CPDs that containcytosine. For energeticreasons, direct absorptionof UVA generates primarilythymine-containing (TT) cyclo-butane pyrimidine dimers. (A)There is no dark CPD formationin albino melanocytes afterUVA exposure.The slope ofpost-UVA CPD induction inalbino cells is indistinguishablefrom the slope in unirradiatedcells. Data represent the aver-age of three experiments,expressed as a fraction of thetotal CPDs generated at0 hours. Hence, the total CPDline has no error bar.The totalnumber of CPDs at0 hours was 169 CPDs permegabase of DNA. (B) DarkCPDs are induced in melanin-containing melanocytes byUVA and include a greaternumber of cytosine-containingTC and CTdimers, capable ofcausing UV-signature C→Tmutations. Slopes for post-UVA induction of TT,TC, and CT CPDs in melanin-containing cells are greaterthan those in albino,withP=0.01,0.05, and 0.03, respectively.The total numberof CPDs at 0 hourswas 87CPDs per megabase of DNA, consistent with the shielding function of eumelanin. Data represent theaverage of five experiments.

Fig. 1. Cyclobutane pyrimidine dimers (CPDs) continue to be gener-ated in melanocytes long after UV exposure ends. (A to C) In albinomurine melanocytes (A), CPD induction at time 0 is followed by repair,but in melanin-containing melanocytes [(B) and (C)], CPDs continue to

increase for 4 hours after UVA exposure ends. CPDs were assayed by DNA ELISA. (D) Additional dark CPDs are revealed in melanocytes when excisionrepair is suppressed by siXpa or siXpc; dark CPDs account for half the total CPDs. (E) In melanocytes from humans, the production of dark CPDs afterUVA varies between individuals. (F) “Nonproducers” are revealed as producers after DNA repair is suppressed in the cells by siXPA or siXPC. (G) DarkCPDs in mouse melanocytes and keratinocytes in vivo. K14-Kitl transgenic mice, which have epidermal melanocytes containing eumelanin, were crossedto mice carrying the Mc1re/e allele, which confers epidermal pheomelanin and yellow fur. Scale bar, 50 mm. (H) Quantitation of CPDs in epidermalsections. Error bars are SD from four experiments [(A) to (C)], three experiments [(D) to (F)], or two experiments (H). P values by t test are for thedifference between the asterisked time point and 0 hours or as indicated. *P ≤ 0.05, **P ≤ 0.005, ***P ≤ 0.0005, ****P ≤ 0.00005.

RESEARCH | RESEARCH ARTICLESon F

ebruary 23, 2021

http://science.sciencemag.org/

Dow

nloaded from

N-acetylcysteine and a-tocopherol (vitamin E).The former suppressed production of dark CPDsby 64% and the latter abolished it (fig. S2, A andB). An effect of antioxidants on CPDs detectableimmediately after UV has been reported, but waslinked to changes in chromatin structure (20).Because melanin generates superoxide radicalion (O2

•–) while it is being irradiated, we specif-ically scavenged it with TEMPOL (1-oxyl-2,2,6,6-tetramethyl-4-hydroxypiperidine) and found thatthis also abolished the production of dark CPDsafter UVA (fig. S2C). A longer-lasting source ofO2

•– in cells is NOX [reduced nicotinamide adeninedinucleotide phosphate (NADPH) oxidase], whichis rapidly induced byUVA (19). TheNOX inhibitorVAS2870 also abolished production of dark CPDs(Fig. 3B). CPDs were inhibited to a level belowthat at 0 hours, indicating that dark CPDs werealso induced during the irradiation itself.The O2

•– requirement recalled an old observa-tion (21) that CPDs can be generated in the darkby dioxetane, a strained four-member ring per-oxide. This moiety undergoes thermolysis to twocarbonyls, with one in an electronically excitedtriplet state having the high energy of aUVphoton(>3 eV, 70 kcal/mol, or 35,000 K) and capable oftransferring its energy to DNA by a process thatis independent of radiation (21). One of the fewbiologically synthesized molecules that can ini-tiate such “photochemistry in the dark” (22, 23)is peroxynitrite (ONOO–) (24), which is gener-ated by the reaction of O2

•–with nitric oxide (NO•).Peroxynitrite and its precursors are stable enoughto diffuse several cell diameters from the site ofgeneration to the target molecule. NOX and the

NO• generator iNOS are UV-inducible withinminutes and are expressed in melanocytes, in-cluding nuclear NOX4 (25). To determine whetherNO• is required for production of dark CPDs, weinhibited iNOS by treating melanocytes withaminoguanidine. This resulted in complete sup-pression of dark CPD production (Fig. 3C).Wenext investigatedwhethermelanin-containing

cells generate peroxynitrite when exposed to UV.Peroxynitrite can be detected by its nitration oftyrosines. Assays for 3-nitrotyrosine revealed thatbasal nitration activity was present even withoutUV, that it was located primarily in the nucleus,and that its appearance requiredmelanin (Fig. 3,D and E, and fig. S2D). Themelanin requirementsuggests that theNO• requirement inmelanocytesis distinct from that observed in cultured keratin-ocytes (18). Peroxynitrite’s presence in the nucleuswithout UV exposure is plausible because the syn-thesis of melanin monomers is a redox reactionthat releases O2

•– (7) and melanosomes are as-sembled in the perinuclear space. In support of anadditional UV role, we found that exposure ofmelanin-containing melanocytes to UVA createdas much nuclear and cytoplasmic nitrotyrosinein 30 min as had accumulated from basal induc-tion over ~8 days (Fig. 3, D and E, and fig. S2D)(14). This represents a factor of ~400 spike in theflux of peroxynitrite per hour.A diagnostic for excited triplet states is that

their energy can also discharge via ultraweakblue or green luminescence (23). This signal canbe enhanced by several orders of magnitude bydiverting the energy to a triplet energy acceptor,sodium 9,10-dibromoanthracene-2-sulfonate (DBAS),

which is a more efficient emitter. To allow rapidDBAS entry, we permeabilizedmelanocytes withina liquid scintillation counter set to single-photoncounting mode. UVA-irradiated melanocytes gen-erated DBAS-amplified luminescence for severalhours after irradiation, and only in cells con-taining melanin (Fig. 3, F and G). The lifetime oftriplet carbonyls is ~10 ms (24), so a signal per-sistent over the course of hours indicates on-going creation. To test whether this triplet stateis required for dark CPDs, we added ethyl sor-bate, a specific quencher of triplet states (26), tointact melanocytes after UVA exposure. We foundthat this treatment prevented production of darkCPDs (Fig. 3H). Because DBAS diverts triplet en-ergy to luminescence, this compound also blockeddark CPDs (fig. S2E). These results provide directevidence that UV-induced superoxide and nitricoxide lead to a high-energy triplet state moiety,which then creates a dark CPD by energy transfer.

Electronically excited melanin asa molecular vector

Melanin is located in the cytoplasm, yet CPDswere generated in the nucleus, raising the ques-tion of what molecule acquired the energetic car-bonyl. An O2

•–-initiated radical chain reaction inlipids was one possible conduit (26), but inducinglipid peroxidation in isolated nuclei by exposureto cumene hydroperoxide did not appear to in-duce CPDs in nuclear DNA (fig. S3).Another candidate for the triplet-state carrier

was the set ofmelaninmonomers out of which thefinalmelaninpolymers are assembled. Thesemono-mers are lipophilic and are therefore potentially

844 20 FEBRUARY 2015 • VOL 347 ISSUE 6224 sciencemag.org SCIENCE

Fig. 3. In the proposedphotochemical pathwayfor production of darkCPDs, high-energy triplet-state moieties are createdby melanin, superoxide,and nitric oxide. DarkCPDs are blocked byinhibitors of (A) melaninsynthesis (kojic acid,KA), (B) the superoxidegenerator NADPH oxidase (VAS2870), and (C) the nitric oxide generatoriNOS (aminoguanidine). Inhibitors are not expected to reduce CPD levelsbelow those seen at 0 hours, a time when CPDs are generated primarily bydirect UV absorption (dotted line). (D) The product of superoxide and nitricoxide, peroxynitrite, generates 3-nitrotyrosine–modified proteins inmelanocytenuclei (except in albino melanocytes), and these increase with UVA exposure.(E) Quantitation of 3-nitrotyrosine in melanin-containing melanocytes. Theincrease at 0 hours arises from protein nitration during the 27-min UVAexposure. (F) Ultraweak chemiluminescence is generated by permeabi-

lized UVA-irradiatedmelanizedmelanocytes that had been incubated with atriplet energy acceptor probe, DBAS.Chemiluminescence was quantified bysingle-photon counting (cpm, counts per minute). (G) Albino melanocytesdo not produce chemiluminescence when irradiated. (H) A triplet-statequencher, ethyl sorbate, inhibits production of dark CPDs in pigmentedmouse melanocytes after they are exposed to UVA. Data are the average ofthree experiments. (F) and (G) show data from one of three similar exper-iments. P values are for the difference between treated and untreatedsamples.

RESEARCH | RESEARCH ARTICLESon F

ebruary 23, 2021

http://science.sciencemag.org/

Dow

nloaded from

able to enter the nucleus, and are concentratedin perinuclear coated vesicles before transfer tomelanosomes (27). Melanin polymer rapidly sol-ubilizes when exposed to hydrogen peroxide(H2O2) (28), and its degradation by peroxida-tion or UV photoionization has been proposed toinvolve dioxetane and triplet carbonyls (29, 30).We first investigated whether a triplet state couldbe created in melanin by a cell-free system. Syn-thetic melanin oxidized with peroxynitrite andincubated with DBAS generated a chemilumines-cence signal two orders of magnitude larger thanthat seen in melanocytes (Fig. 4A and fig. S4A).We found that chemiluminescence began imme-diately and proceeded for >10 min. To furthertest melanin’s ability to host a triplet state, weoxidized synthetic melanin with horseradishperoxidase in the presence of hydrogen peroxide,which generates dioxetane intermediates by amechanism similar to that of peroxynitrite (24).This produced similar levels of chemiluminescence(Fig. 4A and fig. S4B). The eumelanin monomer5,6-dihydroxyindole-2-carboxylic acid (DHICA)and the pheomelaninmonomer5-S-cysteinyldopa(5SCD) also generated chemiluminescence whenoxidized (Fig. 4A).We found that CPDs were generated in the

complete absence of UV when we incubated plas-mid DNA and peroxynitrite withmelanin, DHICA,or 5SCD (Fig. 4B). We were unable to test whether

the triplet quencher ethyl sorbate blocked thesereactions because it was insufficiently soluble inthe aqueous buffer. However, we found that DBAS,which redirects triplet energy toward lumines-cence, reduced CPD production by 50 to 90%(Fig. 4C and fig. S4C). The CPDs created by oxi-dizing melanin or its monomers included themutagenic cytosine-containing CPDs (fig. S4D).The level of CPDs induced in the absence of UVwas approximately equal to that generated in pureDNA by 25 kJ/m2 of UVA—an exposure about one-quarter of that required to produce a barely per-ceptible sunburn (the “minimal erythema dose”).On the basis of our mass spectrometry data fromalbino murine melanocytes (Fig. 2), this value isapproximately 1 CPD per 24 kb of DNA createdsolely by oxidized melanin.We next explored whether UV exposure over-

comes themigration barrier posed by the nuclearmembrane. Cell-free experiments revealed thatmelanin polymer was rapidly solubilized afterexposure to UVA or peroxynitrite (Fig. 4D). Be-cause antibodies to melanin monomers do notexist, we were unable to directly monitor mel-anin migration into the nucleus. We thereforeimaged unfixed melanocytes (to avoid artifac-tually permeabilizing the nuclear membrane),using differential interference contrast micros-copy to maintain contrast at high magnification.Unirradiated normal melanocytes, but not albino

melanocytes, showed dark granules in the cyto-plasm, especially in the perinuclear area (fig. S5).We presume that these are melanin aggregatesin melanosomes, coated vesicles, and endoplas-mic reticulum because they have the same sizeand cellular locations as granules seen in mela-nocytes immunostained for tyrosinase (31). AfterUVA exposure, 3D images reconstructed fromserial planar images revealed granules inside thenucleus (movies S1 and S2). We do not knowwhether these represent melanin-containing or-ganelles that moved into the nucleus or molecularaggregates formed by spontaneous polymeriza-tion of melanin precursors, as occurs during nor-mal melanin synthesis. We conclude that in cellsexposed to UV, melanin or its constituents areable to enter the nucleus.Finally, we sought to identify carbonyl-containing

DHICA reaction products remaining after thetriplet excitation energy discharges. The milderreaction of horseradish peroxidase with hydro-gen peroxide was used to avoid complete oxida-tion and thereby reveal labile reaction intermediates.The DHICA analog of the predicted eumelanindegradation product (29) (Fig. 4E, structure 1)has amolecular weight of 225 daltons and wouldbe highly polar. An alternative reaction is for-mation of a dioxetane moiety at the pyrrole ring(32), also leading to a polar structure of 225 dal-tons (structure 2). When we separated reaction

SCIENCE sciencemag.org 20 FEBRUARY 2015 • VOL 347 ISSUE 6224 845

Fig. 4. Excited-state melanin may act as a molecular vector. (A) Syntheticmelanin, its eumelanin monomer DHICA, and its pheomelanin monomer 5SCDgenerate ultraweak chemiluminescence when oxidized by peroxynitrite or byhorseradish peroxidase plus H2O2. DBAS was used as a triplet energy probe,so luminescence indicates the presence of electronically excited triplet states.Data represent one of six experiments. (B) CPDs are created in the dark whenmelanin or its monomers are oxidized with peroxynitrite. The level of CPDsresulting from direct DNA absorption of UVA is shown for comparison. (C)Diverting triplet energy to luminescence with DBAS blocks production of CPDsby oxidized melanin.The UVC dose used for comparison is about two orders ofmagnitude higher than that lethal to 50% of normal cells. In (B) and (C), data

represent the average of three experiments. P values are for the difference be-tween treated and “DNA only” samples or as indicated. (D) Synthetic melanin israpidly solubilized upon exposure to UVor peroxynitrite. Left to right: Supernatantsresulting from untreated melanin or from melanin exposed to UVA (200 kJ/m2),7.8 mM NaOH, or peroxynitrite (1 mM in NaOH stabilizer). (E and F) Identificationof a putative triplet-carbonyl carrier. (E) The eumelanin monomer DHICA wasoxidized and products were separated by HPLC, monitoring the HPLC outputby mass spectrometry in negative ion mode. Monitoring for the 224m/z ionof 225-dalton compounds 1 and 2, expected for triplet-carbonyl derivativesof dioxetane-adducted DHICA, reveals a single polar peak. (F) Scanning thepolar HPLC peak reveals four principal m/z peaks, one at 224 m/z.

RESEARCH | RESEARCH ARTICLESon F

ebruary 23, 2021

http://science.sciencemag.org/

Dow

nloaded from

products by high-performance liquid chromato-graphy (HPLC) and scanned the HPLC eluate bymass spectrometry in negative ion mode for thecorresponding 224 m/z product, we observed asingle peak eluting in the highly polar region ofthe buffer gradient (Fig. 4E). Scanning this frac-tion of the HPLC eluate across the small-moleculeregion of the mass spectrometer revealed fourprincipalm/z peaks, of which the second largestwas 224 daltons (Fig. 4F). Fragmentation of theseHPLC fractions generated m/z peaks assignableto oxidation products of either structure 1 or 2created by known oxidation reactions. When weblocked the reactivity of DHICA’s six-memberedring by omitting an OH group (5-hydroxyindole-

2-carboxylic acid) (fig. S6A), oxidation still pro-duced chemiluminescence (fig. S6, B and C) andCPDs (fig. S6D), which suggests that 2 is thepredominant product. Together, these results in-dicate that melanin and melanin fragments arecapable of acting as the molecular vectors thatacquire an electronically excited triplet state,probably at a carbonyl arising from a dioxetaneintermediate.

Proposed scheme for the participation ofmelanin in melanoma development

We have shown that exposing melanin-containingcells to UV radiation induces superoxide and nitricoxide, causing a factor of ~400 peroxynitrite spike

that degrades melanin, allows melanin-like gran-ules to appear in the nucleus and, for hours afterthe original UV exposure, excites melanin deriva-tives to a triplet state that has the high energy of aUV photon (Fig. 5). These evanescent electronicallyexcited products transfer their triplet energy toDNA, creating mutagenic cyclobutane pyrimi-dine dimers in the dark. We speculate that thedegradation of melanin polymer into fragmentsallows thesemoieties to closely approach theDNA,and that the peroxynitrite-melanin reaction inter-mediate is a short-lived dioxetane (fig. S6E)whosetriplet energy level lies well above that of theDNA bases. The sustained time course of darkCPD generation can be accounted for by the pro-longed steps of UV induction of NOX and iNOS,peroxynitrite-induced solubilization ofmelanin tofragments (or release of pre-melanin monomersfrom UV-damaged melanosomes and vesicles),and migration to the nucleus. Still unidentifiedare the isozymes generating superoxide andnitric oxide, the cell site(s) at which melanin isdegraded and its fragments are excited to a trip-let state, the full inventory of eumelanin andpheomelanin fragments that can host tripletstates, the chemical intermediate through whichONOO– creates the triplet carbonyl, and the en-ergy transfer process.A consequence of these events is that melanin

may be carcinogenic as well as protective againstcancer. This double nature would explain the ap-parent cancer-facilitating effects of melanin seenin mice and in human epidemiology (2–4, 6).Melanin is an unusual polymer whose propertiesset the stage for the events we have described(33). Highly reactive o-quinones created by ROS-generating redox transformations of tyrosinepolymerize spontaneously into oligomers. Theo-quinones readily accept an electron to becomesemiquinone radicals, giving melanin a highconcentration of free radicals in stable redoxequilibrium and stabilized by metal ions. Thismacromolecule is a photon trap that also actsas an electron-proton photoconductor. Thesecharacteristics give melanin its broad light ab-sorption, radical scavenging, and metal reservoirproperties, but at a price. First,melanin synthesisgenerates O2

•– and H2O2. UV exposure addition-ally excites the rings to an energy that, especiallyfor pheomelanin, ejects an electron that is cap-tured by oxygen to yieldmore O2

•– (5, 34). Second,the reactive semiquinones allow melanin to bedegraded and these fragments to be adductedto create high-energy unstable moieties such asdioxetanes. Althoughmost of the cell’s melaninsynthesis is safely isolated inside melanosomes,the early steps occur in close proximity to thenucleus.It was proposed long ago that chemiexcitation—

the creation of chemical reaction products con-taining excited electrons that underlie biolumi-nescence in lower organisms—has broad importancein biology (22, 23). Our data suggest that thismay be the case in human skin. The consequenceis that half or more of the CPDs in a melanocytearise after UV exposure ends. In vivo the sameappears to be true of keratinocytes, which receive

846 20 FEBRUARY 2015 • VOL 347 ISSUE 6224 sciencemag.org SCIENCE

Fig. 5. A mechanistic modelfor the generation of darkCPD in melanocytes bychemiexcitation, with mela-nin as an active participant.Exposing cells to UV radiationis known to up-regulate iNOS,NADPH oxidase (NOX), andenzymes of melanin synthesis,presumably causing sustainedgeneration of nitric oxide (NO•)and superoxide (O2

•–). Cyto-plasmic NOS and NOX areindicated on the figure butsome isoforms are located inthe nucleus. The presentexperiments show a UV-induced surge in the productof these two radicals, thepowerful oxidant peroxynitrite(ONOO–), and show that per-oxynitrite degrades melaninpolymer to fragments. Melaninor melanin fragments thenappear in the nucleus. Perox-ynitrite is also one of the fewbiologically synthesized mole-cules capable of exciting anelectron to a triplet state. Thepresent experiments show thaton a faster time scale, perox-ynitrite excites an electron in amelanin fragment to a tripletstate that has the high energyof a UV photon. The typicaltriplet-state reactionintermediate, not demon-strated here (hence indicatedin italics), is a cycloaddition of–O–O– to create an unstabledioxetane; dioxetanes undergospontaneous thermolysis toyield two carbonyls, one ofwhich acquires most of theenergy and finishes in a high-energy triplet state (*). For the melanin-related triplet, the half-life of thereaction intermediate appeared to be minutes, and a carbonyl consistent with a dioxetane precursor wasidentified by mass spectrometry.Triplet energy then discharges on a microsecond time scale to generatevisible luminescence, or discharges in a radiation-independent manner to DBAS (to be emitted as fluo-rescence), to sorbate (to be dissipated as isomerization and heat), or evidently to DNA bases (where itmakes CPDs).The presence of melanin, activation of iNOS and NOX, and the triplet state were shown tobe required for dark CPD formation.

RESEARCH | RESEARCH ARTICLESon F

ebruary 23, 2021

http://science.sciencemag.org/

Dow

nloaded from

melanosomes donated bymelanocytes. If the sameholds for human skin, this would mean thatpast measurements of CPDs immediately afterUV exposure have underestimated the con-sequences of UV exposure.One benefit of dark photochemistry’s slow

course is that it allows intervention. A blockerof dark CPDs, a-tocopherol (vitamin E), is notonly an antioxidant but also inactivates dioxe-tanes by converting them to a pair of diols (35).The triplet quencher ethyl sorbate is an analogof the widely used food preservative potassiumsorbate. Screening for novel triplet quenchersoffers the prospect of developing “evening-after”sunscreens that could potentially prevent thecarcinogenic processes occurring in the skinhours after sunlight exposure ends.

REFERENCES AND NOTES

1. N. Kollias, R. M. Sayre, L. Zeise, M. R. Chedekel, J. Photochem.Photobiol. B 9, 135–160 (1991).

2. P. F. Williams, C. M. Olsen, N. K. Hayward, D. C. Whiteman, Int.J. Cancer 129, 1730–1740 (2011).

3. F. P. Noonan et al., Nat. Commun. 3, 884 (2012).4. D. Mitra et al., Nature 491, 449–453 (2012).5. M. R. Chedekel, S. K. Smith, P. W. Post, A. Pokora,

D. L. Vessell, Proc. Natl. Acad. Sci. U.S.A. 75, 5395–5399(1978).

6. S. Takeuchi et al., Proc. Natl. Acad. Sci. U.S.A. 101,15076–15081 (2004).

7. J. L. Munoz-Munoz et al., Biochim. Biophys. Acta 1794,1017–1029 (2009).

8. D. E. Brash, Photochem. Photobiol. 91, 15–26 (2015).9. M. Krauthammer et al., Nat. Genet. 44, 1006–1014

(2012).10. D. E. Brash, S. Seetharam, K. H. Kraemer, M. M. Seidman,

A. Bredberg, Proc. Natl. Acad. Sci. U.S.A. 84, 3782–3786(1987).

11. W. J. Schreier et al., Science 315, 625–629 (2007).12. D. Lazovich et al., Cancer Epidemiol. Biomarkers Prev. 19,

1557–1568 (2010).13. T. Douki, A. Reynaud-Angelin, J. Cadet, E. Sage, Biochemistry

42, 9221–9226 (2003).14. See supplementary materials on Science Online.15. K. H. Kraemer et al., Proc. Natl. Acad. Sci. U.S.A. 72, 59–63

(1975).16. W. Carrier, R. Snyder, J. Regan, in The Science of

Photomedicine, J. Regan, J. Parrish, Eds. (Plenum, New York,1982), pp. 91–112.

17. E. J. Beecham, J. F. Mushinski, E. Shacter, M. Potter, V. A. Bohr,Mol. Cell. Biol. 11, 3095–3104 (1991).

18. C. Gordon-Thomson et al., Photochem. Photobiol. Sci. 11,1837–1847 (2012).

19. A. Valencia, I. E. Kochevar, J. Invest. Dermatol. 128, 214–222(2008).

20. M. Hochberg, R. Kohen, C. D. Enk, Biomed. Pharmacother. 60,233–237 (2006).

21. A. A. Lamola, Biochem. Biophys. Res. Commun. 43, 893–898(1971).

22. E. White, J. Miano, C. Watkins, E. Breaux, Angew. Chem. Int. Ed.Engl. 13, 229–243 (1974).

23. G. Cilento, in Chemical and Biological Generation of ExcitedStates, W. Adam, G. Cilento, Eds. (Academic Press, New York,1982), pp. 277–307.

24. F. S. Knudsen et al., Chem. Res. Toxicol. 13, 317–326(2000).

25. C. Roméro-Graillet et al., J. Biol. Chem. 271, 28052–28056(1996).

26. A. C. Velosa, W. J. Baader, C. V. Stevani, C. M. Mano,E. J. Bechara, Chem. Res. Toxicol. 20, 1162–1169 (2007).

27. S. Hatta, Y. Mishima, M. Ichihashi, S. Ito, J. Invest. Dermatol.91, 181–184 (1988).

28. P. Kayatz et al., Invest. Ophthalmol. Vis. Sci. 42, 241–246(2001).

29. D. Slawinska, J. Slawinski, Physiol. Chem. Phys. 14, 363–374(1982).

30. K. Wakamatsu, Y. Nakanishi, N. Miyazaki, L. Kolbe, S. Ito,Pigment Cell Melanoma Res. 25, 434–445 (2012).

31. R. Halaban et al., Proc. Natl. Acad. Sci. U.S.A. 97, 5889–5894(2000).

32. I. Saito, S. Matsugo, T. Matsuura, J. Am. Chem. Soc. 101,4757–4759 (1979).

33. P. Meredith, T. Sarna, Pigment Cell Res. 19, 572–594(2006).

34. T. Ye et al., Photochem. Photobiol. 82, 733–737(2006).

35. W. Adam, F. Vargas, B. Epe, D. Schiffmann, D. Wild, Free Radic.Res. Commun. 5, 253–258 (1989).

ACKNOWLEDGMENTS

We thank M. Bosenberg and V. Muthusamy for UVA-irradiatedmouse skin; the Yale Office of Environmental Health and Safetyfor the single-photon liquid scintillation counter; A. Bommakantifor photography; and D. Mitchell and A. Mennone for helpfuldiscussions. Supported by Department of Defense CDMRP grantsCA093473P1 and CA093473 (D.E.B. and R.H.); NIH grant 2 P50

CA121974 (R.H. and D.E.B.); a postgraduate fellowship from theUniversity of Veterinary Medicine, Vienna (S.W.); Fundação deAmparo a Pesquisa do Estado de São Paulo (FAPESP) doctoralfellowship 09/02062-8 (C.M.M.); and FAPESP grant 06/56530-4and the INCT–Processos Redox em Biomedicina (E.J.H.B.).Confocal support was provided by NIH grant P30 DK34989 tothe Yale Liver Center.

SUPPLEMENTARY MATERIALS

www.sciencemag.org/content/347/6224/842/suppl/DC1Materials and MethodsSupplementary TextFigs. S1 to S6Movies S1 and S2References (36–39)

13 May 2014; accepted 2 January 201510.1126/science.1256022

STRUCTURAL BIOLOGY

Structural basis for Notch1engagement of Delta-like 4Vincent C. Luca,1,2,3 Kevin M. Jude,1,2,3 Nathan W. Pierce,2 Maxence V. Nachury,2

Suzanne Fischer,1,2,3 K. Christopher Garcia1,2,3*

Notch receptors guide mammalian cell fate decisions by engaging the proteins Jaggedand Delta-like (DLL). The 2.3 angstrom resolution crystal structure of the interactingregions of the Notch1-DLL4 complex reveals a two-site, antiparallel binding orientationassisted by Notch1 O-linked glycosylation. Notch1 epidermal growth factor–like repeats11 and 12 interact with the DLL4 Delta/Serrate/Lag-2 (DSL) domain and module at theN-terminus of Notch ligands (MNNL) domains, respectively. Threonine and serineresidues on Notch1 are functionalized with O-fucose and O-glucose, which act as surrogateamino acids by making specific, and essential, contacts to residues on DLL4. Theelucidation of a direct chemical role for O-glycans in Notch1 ligand engagementdemonstrates how, by relying on posttranslational modifications of their ligand bindingsites, Notch proteins have linked their functional capacity to developmentallyregulated biosynthetic pathways.

The Notch signaling pathway is essential forcell-fate determination in all metazoan spe-cies (1). In adult mammals, Notch signalingdirects neural and hematopoietic stem celldifferentiation and development of many

immune cell subsets (2–4). Mammalian Notch re-ceptor homologs (Notch1 to 4) encode a Notchextracellular domain (NECD) that engages ligands,a transmembrane domain, and a Notch intracel-lular domain (NICD) that translocates to thenucleus to serve as a transcriptional cofactor(5). Signaling is initiated when the NECD bindsJagged1, Jagged2, Delta-like 1 (DLL1), or Delta-like 4 (DLL4) ligands on the surface of an apposingcell, triggering proteolysis of the Notch receptorin a process that is dependent on ligand endo-cytosis (6–9). Mutations in Notch and its ligandsare associated with a number of inherited and

acquired diseases, including Alagille syndrome(AGS) and T cell acute lymphoblastic leukemia(T-ALL) (10).Notch-ligand interactions are apparently de-

pendent on posttranslational modification ofNotch receptors with O-glucose (O-Glc) and O-fucose (O-Fuc). The epidermal growth factor (EGF)modules of the NECD are specifically targetedby protein O-glucosyltransferase 1 (Poglut) andprotein O-fucosyltransferase 1 (Pofut1) enzymes(11, 12). O-glucosylation and O-fucosylation areboth required for Notch activation yet play dis-tinct roles in the signaling pathway. O-glucosemodifications indirectly affect signaling by in-creasing susceptibility to protease processing afterligand engagement, whereas O-fucose modifica-tions directly enhance Notch affinity for Jagged/DLL (11–13). Elongation of O-fucose to a disaccha-ride by Fringe N-acetyl-glucosaminyltransferasesfurther influences specificity by facilitating pref-erential interactions between certain receptor-ligand pairs (14–16). A long-standing questionhas been whether O-glycans directly contactligands or contribute through allosteric effects.There are few examples of posttranslational

SCIENCE sciencemag.org 20 FEBRUARY 2015 • VOL 347 ISSUE 6224 847

1Howard Hughes Medical Institute, Stanford UniversitySchool of Medicine, Stanford, CA 94305, USA. 2Departmentof Molecular and Cellular Physiology, Stanford UniversitySchool of Medicine, Stanford, CA 94305, USA. 3Departmentof Structural Biology, Stanford University School of Medicine,Stanford, CA 94305, USA.*Corresponding author. E-mail: kcgarcia@stanford.edu

RESEARCH | RESEARCH ARTICLESon F

ebruary 23, 2021

http://science.sciencemag.org/

Dow

nloaded from

Chemiexcitation of melanin derivatives induces DNA photoproducts long after UV exposure

Bechara, Ruth Halaban, Thierry Douki and Douglas E. BrashSanjay Premi, Silvia Wallisch, Camila M. Mano, Adam B. Weiner, Antonella Bacchiocchi, Kazumasa Wakamatsu, Etelvino J. H.

DOI: 10.1126/science.1256022 (6224), 842-847.347Science 

, this issue p. 842; see also p. 824Scienceenergy could be dissipated by adding quenchers to sunscreens.energy is then transferred to DNA, inducing the same damage as ultraviolet light, but in the dark. Conceivably, this which reactive oxygen and nitrogen species induced by ultraviolet light excite an electron in melanin fragments. Thispigment thought to protect against cancer (see the Perspective by Taylor). They propose a ''chemiexcitation'' model in

find that a key mediator of this delayed damage is melanin, aet al.generated for hours after sunlight exposure. Premi beach. It seems that the DNA photoproducts responsible for cancer-causing mutations in skin cells continue to be

Sun worshippers may have more to worry about than the DNA damage that occurs while they're relaxing on theThe dark side of melanin exposed

ARTICLE TOOLS http://science.sciencemag.org/content/347/6224/842

MATERIALSSUPPLEMENTARY http://science.sciencemag.org/content/suppl/2015/02/18/347.6224.842.DC1

CONTENTRELATED

file:/contenthttp://science.sciencemag.org/content/sci/346/6212/934.fullhttp://stm.sciencemag.org/content/scitransmed/5/196/196ra98.fullhttp://stm.sciencemag.org/content/scitransmed/5/200/200ra116.fullhttp://stm.sciencemag.org/content/scitransmed/5/202/202ra123.fullhttp://stm.sciencemag.org/content/scitransmed/6/254/254ra128.fullhttp://science.sciencemag.org/content/sci/347/6224/824.full

REFERENCES

http://science.sciencemag.org/content/347/6224/842#BIBLThis article cites 35 articles, 11 of which you can access for free

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Terms of ServiceUse of this article is subject to the

is a registered trademark of AAAS.ScienceScience, 1200 New York Avenue NW, Washington, DC 20005. The title (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement ofScience

Copyright © 2015, American Association for the Advancement of Science

on February 23, 2021

http://science.sciencem

ag.org/D

ownloaded from