Dithiothreitol (DTT) Acts as a Specific, UV-inducible Cross-linker in Elucidation of Protein–RNA Interactions*

Protein–RNA cross-linking by UV irradiation at 254 nm wavelength has been established as an unbiased method to identify proteins in direct contact with RNA, and has been successfully applied to investigate the spatial arrangement of protein and RNA in large macromolecular assemblies, e.g. ribonucleoprotein-complex particles (RNPs). The mass spectrometric analysis of such peptide-RNA cross-links provides high resolution structural data to the point of mapping protein–RNA interactions to specific peptides or even amino acids. However, the approach suffers from the low yield of cross-linking products, which can be addressed by improving enrichment and analysis methods. In the present article, we introduce dithiothreitol (DTT) as a potent protein–RNA cross-linker. In order to evaluate the efficiency and specificity of DTT, we used two systems, a small synthetic peptide from smB protein incubated with U1 snRNA oligonucleotide and native ribonucleoprotein complexes from S. cerevisiae. Our results unambiguously show that DTT covalently participates in cysteine-uracil crosslinks, which is observable as a mass increment of 151.9966 Da (C4H8S2O2) upon mass spectrometric analysis. DTT presents advantages for cross-linking of cysteine containing regions of proteins. This is evidenced by comparison to experiments where (tris(2-carboxyethyl)phosphine) is used as reducing agent, and significantly less cross-links encompassing cysteine residues are found. We further propose insertion of DTT between the cysteine and uracil reactive sites as the most probable structure of the cross-linking products.

nucleobases to UV light (11,12). The site of cross-linking can then be determined by mass spectrometric analysis (including gas phase fragmentation of the corresponding peptide-nucleic acid conjugates) and database searching. To enhance the low yield of protein-DNA cross-linking, derivatives with higher UV reactivity, such as halonucleotides, can be employed (13)(14)(15). UV-induced cross-linking is more frequently used to monitor protein-RNA interactions. RNA is mainly present in its single-stranded form, and adopts a variety of tertiary structures in which the bases of the nucleotides are in close contact with amino acid residues of the proteins that are bound to the RNA. Several studies have used this technique to identify globally the proteins that directly interact with e.g. poly(A) mRNA in yeast and human systems, and have yielded insight into the large variety of RNA-binding proteins that exist within the cell (16 -21). More detailed approaches apply UVinduced protein-RNA cross-linking in a similar manner, but extend the MS analysis toward the identification of the crosslinked amino acids together with the corresponding nucleotide moieties, allowing exact definition of the RNA-binding regions in the cross-linked proteins (22)(23)(24). To improve the yield of cross-linking, more reactive nucleoside derivatives (4-thiouridine and 6-thioguanosine) have been incorporated into RNA in growing cells. The cross-linking reaction of these derivatives with amino acids is not entirely additive, however, but is accompanied by the loss of H 2 S ((22), U.Z. and H.U, unpublished results). Other cross-linking reactions between proteins and RNA have made use of nucleotide derivatives that contain a cross-linkable function at the 2Ј hydroxyl group or the phosphate moiety (25)(26)(27), but have not been characterized by MS yet.
Additional cross-linking agents for the analysis of protein-RNA interactions have been exploited, but have not yet found their way into modern MS-based proteome analyses. For instance, methylene blue has been described as a light-inducible cross-linker, in particular for mapping interactions of proteins with dsRNA (28). Similarly, protein-RNA interaction studies in ribosomal subunits have made use of diepoxybutane, or nitrogen mustard (29 -31). The same holds true for the use of 2-iminothiolane ("Traut's reagent"), which is a protein-RNA cross-linking reagent that combines chemical and UVinducible features and has been extensively applied to the analysis of protein-RNA contacts in ribosomal subunits (33,32). Here, we introduce dithiothreitol (DTT) 1 as a potent UVinducible cross-linking reagent for the analysis of protein-RNA linkages following UV irradiation. By exhaustive massspectrometric analyses we found that upon UV light exposure DTT forms a covalent linkage between cysteine residues within proteins and uracil bases of RNA in close proximity. We applied this to protein-RNA complexes isolated from yeast cells and compared the protein-RNA cross-linking patterns that were obtained in the presence and absence of DTT. We found that the cross-linking reaction is surprisingly efficient and specific.

EXPERIMENTAL PROCEDURES
In Vitro Transcription of 13 C-Labeled RNA Oligonucleotides-Human U4 snRNA 5Ј stem loop oligonucleotide GCCAAUGAG-GUUUAUCCGAGGC (position 26 -47) was synthesized by in vitro transcription with T7 RNA polymerase (34). DNA template with T7 promoter sequence was obtained from Eurofins (Eurofins MWG Operon Ebersberg Germany). Annealing of DNA template (1 M) with T7 promoter (1 M) was performed by heating at 90°C for 2 min. Four 13 C labeled oligonucleotides were synthesized, each with one labeled ribonucleoside triphosphate (rNTP) and three unlabeled rNTPs, and final concentrations of 1 mM and 4 mM, respectively. Unlabeled rNTPs (Jena Bioscience, Jena, Germany) and 13 C labeled rNTPs (Sigma-Aldrich, Taufkirchen, Germany) were obtained as sodium salts. Transcription was performed in 40 mM Tris-HCl pH 8.0, 10 mM DTT, 2 mM spermidine, 30 mM MgCl 2 , and 0.1 mg/ml T7 RNA polymerase. After incubation at 37°C for 4 h the transcription product was purified using 20% polyacrylamide gels (0.7 mm, 20 ϫ 30 cm) containing 7 M urea. RNA oligonucleotide bands on the gels were visualized by UV shadowing, excised and eluted using TEN buffer (10 mM Tris-HCl pH 8.0, 1 mM EDTA, 300 mM NaCl). Finally oligonucleotides were recovered by ethanol precipitation, dissolved in water and stored at Ϫ20°C. Concentration was determined by UV absorption at 260 nm.
UV Cross-linking with Isotope-labeled RNA Oligonucleotides-Recombinant 15.5K protein was purified from Escherichia coli BL21 cells as described by Nottrott et al. (35). For UV cross-linking 1 nmol of 13 C labeled U4 snRNA 5Ј stem loop oligonucleotide was incubated with 1 nmol of 15.5K protein in buffer (20 mM HEPES, 50 mM NaCl, 1.5 mM MgCl 2 ) on ice for 30 min. UV-irradiation was performed in microtiter plates at 254 nm for 10 min. After ethanol precipitation the sample was dissolved with 1 M urea in 50 mM Tris-HCl pH 7.9. The RNA moiety was hydrolyzed by addition of 1 g of RNase A and 1 unit of RNase T1 for 2 h at 52°C. Protein was digested with trypsin (protein/ enzyme 50:1, w/w) overnight at 37°C. Removal of noncross-linked RNA was achieved by solid phase extraction using reversed phase C18 microcolumns and cross-linked peptides enriched using TiO 2 microcolumns, both prepared in-house (36). Noncross-linked controls were prepared and analyzed in parallel.
Samples were loaded onto a nanoflow liquid chromatography system (Agilent 1100 series, Agilent Technologies, Bö blingen, Germany) coupled to an LTQ-Orbitrap Velos mass spectrometer equipped with a FlexIon nanoSpray source (Thermo Fisher Scientific, Bremen, Germany). The chromatography system was operated with a vented column setup (37) at an analytical flow rate of 300 nl/min achieved through passive splitting. Samples were desalted on a self-packed reversed phase-C18 pre-column (20 mm length, 0.15 mm inner diameter, ReproSil-Pur C 18 -AQ 5 m resin, Dr. Maisch GmbH, Ammerbuch-Entringen, Germany). Separation was achieved on a selfpacked reversed phase-C18 analytical column (150 mm length, 0.075 mm inner diameter, ReproSil-Pur C 18 -AQ 3 m resin, Dr. Maisch GmbH) packed into a SilicaTip emitter (FS360 -75-10-N, New Objective, Woburn, MA). Analytes were separated using a linear gradient of 3-36% acetonitrile over 37 min using 0.1% formic acid as modifier, 1 The abbreviations used are: DTT, dithiothreitol; RNPs, ribonucleoprotein complexes; rNTP, ribonucleoside triphosphate; 13 C-rNTP, 13 C labelled ribonucleoside triphosphate; d 10 -DTT, perdeuterated dithiothreitol; TCEP, (tris(2-carboxyethyl)phosphine); HCD, higher energy collision-induced dissociation; MS/MS, product ion scan mass spectrum; FWHM, full width half maximum; 13 UTP, 13 C labelled uridine triphosphate; 13 C-A, 13 C-G, 13 C-U 13 C-C, 13 C labelled adenine, guanine, uracil and cytosine. followed by an increase to 95% acetonitrile versus 0.1% formic acid over 0.5 min and a 7.5 min hold at these conditions. MS data was acquired in positive ion mode using a data-dependent top10 method dynamically choosing the most abundant precursor ions for higher energy collision-induced dissociation (HCD). MS survey spectra were acquired across the 350 -1600 m/z range at a resolution setting of 30,000 full width half maximum (FWHM). MS/MS spectra were acquired at a resolution setting of 7500 FWHM with a normalized collision energy setting of 45%. A minimum of 5000 ion counts were required for triggering fragmentation events.
Analytes were separated using a linear gradient of 4 -36% acetonitrile over 22 min using 0.1% formic acid as modifier, followed by an increase to 95% acetonitrile versus 0.1% formic acid over 0.5 min and a 7.5 min hold at these conditions. A flow rate of 250 nl/min was used.
Data-dependent acquisition was performed in positive ion mode using a top15 method for analysis. MS survey spectra were acquired at a resolution setting of 70,000 FWHM across the 350 -1600 m/z range. HCD with a normalized collision energy setting of 30% was applied for peptide fragmentation, and MS/MS spectra were acquired at a resolution setting of 17,500 FWHM. A minimum of 10,000 ion counts was required for triggering fragmentation events.
UV Cross-linking of Native Ribonucleoprotein Complex-Saccharomyces cerevisiae cells carrying a C-terminal TAP-tag on cbp20 were grown in YPD media (1% yeast extract, 2% bacto-peptone, 2% glucose) to an OD 600 of 5.7 in a fermenter. Cells were harvested and lysate was prepared as described by Kramer et al. (22). TAP-Tag purification was done in two parallel experiments A and B with buffer containing either dithiothreitol (DTT) or TCEP. For both experiments the cell lysate (350 mg protein in 10 ml) was incubated with 600 l of IgG Sepharose TM 6 Fast Flow beads (GE Healthcare, Mü nchen, Germany) equilibrated with AGK buffer (10 mM HEPES, pH 7.5, 1.5 mM MgCl 2 , 200 mM KCl, 10% glycerol, EDTA-free protease inhibitor mixture tablet (Roche, Indianapolis, IN)). RNPs were bound to IgG beads by rotation for 2 h at 4°C. After the flow through was collected, beads were washed with 5 ml of AGK buffer and equilibrated with 20 ml of CBB1 buffer (25 mM Tris-HCl pH 7.9, 150 mM sodium chloride, 1.5 mM magnesium acetate, 2 mM calcium chloride, 1 mM imidazole, 2 mM DTT) in experiment A and with CBB2 buffer (25 mM Tris-HCl pH 7.9, 150 mM sodium chloride, 1.5 mM magnesium acetate, 2 mM calcium chloride, 1 mM imidazole, 2 mM TCEP) in experiment B. RNPs were eluted from IgG by incubating with 12 l PreScission protease and 1 l RNasin overnight at 4°C.
For UV cross-linking, purified RNPs from experiments A and B were divided into control and test samples containing 1 mg of protein each, respectively. Both test samples were UV-irradiated at 254 nm for 2 min on ice in Petri dishes (3.5 cm diameter). After cold ethanol precipitation, protein therein was digested with trypsin (Promega, protein/enzyme 50:1, w/w) overnight at 37°C. Intact RNA and peptide-RNA cross-links were separated from the excess of noncrosslinked peptides by size exclusion chromatography on a Superdex 200 column (3.2 mm ϫ 300 mm) coupled to a SMART chromatography system (GE Healthcare; Uppsala, Sweden) with buffer containing 20 mM Tris pH 7.5, 150 mM NaCl, 1.5 mM MgCl 2 . Noncross-linked RNA was removed by reversed phase-C18 chromatography using inhouse prepared columns (Dr. Maisch, GmbH) following hydrolysis with benzonase for 1 h at 37°C and ribonucleases A/T1 (Ambion, Applied Biosystems, Darmstadt, Germany) as above for 2 h at 52°C. Nonirradiated controls of experiments A and B were processed in the same manner.
LC/MS/MS analysis was performed in positive ion mode using nanoflow liquid chromatography coupled to an LTQ-Orbitrap Velos mass spectrometer with the setup described before, but with a gradient length of 118 min.
MS Data Analysis-Data analysis for identification of putative cross-links was performed by the novel RNP XL pipeline (22) integrated in the OpenMS software environment (38,39). Briefly, the first step is the conversion of raw data (in *.raw format) to the mzML format using the ProteoWizard tool msconvert (40). Two different filters pipelines integrated in the RNP XL pipeline are applied: (1) the ID filter removes MS/MS spectra that match to noncross-linked peptides identified at high confidence by a database search with OMSSA (Open Mass Spectrometry Search Algorithm) (41). (2) the XIC filter reduces false positives by removing the MS/MS spectra of precursors from the UV irradiated sample if they are also present in the nonirradiated control at comparable intensity. The reduced data file is then submitted for generation of precursor mass variants that are obtained by subtracting all potential masses of cross-linked RNA moieties from the observed precursor masses in a combinatorial fashion. This list of precursor mass variants is then submitted for database search by OMSSA. The final output (in idXML format), which contains the true positive hits for cross-linked peptides, is used to annotate raw MS/MS data in TOPPView (42). For a detailed description of data analysis, we refer to the original publication by Kramer et al. and the corresponding tutorial (22).

RESULTS
Working Hypothesis-In recent years we have established a purification strategy that allows for the isolation and subsequent mass spectrometric characterization of cross-linked peptide-RNA oligonucleotides derived from UV-irradiated protein-RNA particles of any kind (22). We were able to sequence a multitude of cross-linked peptides and identified their cross-linked amino acid and nucleotide parts, respectively. In most cases the determined cross-link was found to be of an additive nature, i.e. the mass of the entire cross-link is composed of the mass of its peptide moiety and its RNA moiety. Exceptions to this are observed when the RNA contains modified nucleosides, e.g. 4-thiouridine (44) or 6-thioguanosine (U.Z. and H.U., unpublished results). Another more frequently observed exception is the appearance of peptide-RNA cross-links with a nominal mass increase of 152 Da over the expected additive masses upon UV cross-linking with nonmodified RNAs. We first observed such cross-links in studies of the import factor Snurportin 1 bound to U1 snRNA (45). The peptide 346 GSSHSPDHPGCLMEN 360 derived from Snurportin 1 was found to be cross-linked to an oligonucleotide AAAU (position 102-105 in stem-loop III of U1 snRNA) plus the mass increment of 152 Da. MALDI-MS and ESI-MS/MS experiments revealed that the adduct causing the mass increment is directly attached to the peptide moiety of the cross-link (44).
We detected corresponding adducts upon cross-linking of other recombinant protein-RNA complexes, i.e. in cross-linking studies of the RNase H domain of the human spliceosomal hPrp8 with U4 snRNA (46), the spliceosome associated protein CWC2 with U6 and U4 snRNA (47), but also in crosslinked endogenous protein-RNA complexes where several ribosomal proteins from yeast were found to be cross-linked to RNA with the same mass increment of 152 Da (22). A common feature of all these modified cross-links is that cysteine can be identified as the cross-linked amino acid of the peptide sequence (22).
A comparison of the exact masses detected by high resolution mass spectrometry with the expected additive crosslink masses revealed an exact mass increment of 151.9966 Da (Table I) (Table I), the most plausible ones being C 2 H 5 O 4 N 2 P, C 4 O 3 N 4 , and C 4 H 8 O 2 S 2 . This led us to investigate two hypotheses for the origin of the mass increment: a derivative of a nucleobase formed e.g. by the irradiation used for crosslinking, or a small organic molecule from the pool of reagents used for isolation, purification and cross-linking of the complex. Because the 152 mass increment was observed neither in nonirradiated control samples nor in irradiated TCEP samples (see below), we did not pursue the third probability of a purely proteinaceous origin further.
We set out to evaluate both these two working hypotheses in more detail.
1. UV-induced RNA-RNA cross-linking is a well-known phenomenon that can be exploited for elucidation of RNA-RNA interactions (48,49). It is therefore conceivable that besides the covalent bond formed between cysteine and uracil upon UV cross-linking, another proximal nucleobase or its derivative might form an additional bond to the uracil moiety. To test this hypothesis we used a binary complex that consists of the human spliceosomal U4/U6 snRNP specific protein 15.5K bound to an RNA oligonucleotide harboring the canonical binding site of the 15.5K protein (35). Human 15.5K protein as part of U4/U6.U5 tri-snRNPs was found to be involved in the late stage of spliceosome assembly by binding to the 5Ј stem loop of U4 snRNA (35). The used 22mer RNA oligonucleotide encompasses the U4 snRNA nucleotide positions 26 -47 and was in vitro transcribed with 12 C-rNTPs as well as with all four 13 C-rNTP. If, upon UV cross-linking, the mass increment was induced by a derivative of a nucleobase, we expected to observe a shift of the 152 Da mass increment in the MS analysis corresponding to the incorporation of 13 C-labeled nucleobases. 2. We also evaluated if the mass increment might be because of a small organic molecule either inserted in between, or added to the cysteine thiol and uracil moieties involved in the cross-links. The nominal mass increment of 152 Da e.g. corresponds to DTT, a reducing agent frequently used in cell lysis and subsequent protein purification protocols. To test this hypothesis UV cross-linking was performed with and without DTT, with TCEP and with per- deuterated DTT. For this purpose different experimental models were used. In initial studies a short synthetic peptide derived from U snRNP-specific SmB'/B protein and Sm site 9mer U1 snRNA oligonucleotide was used. Moreover cross-linking was performed in native RNP purified from S. cerevisiae in the presence of either DTT or TCEP.
Evaluation of Nucleoside Derivatives as the Origin of Unusual Mass Adducts-To assess our first hypothesis, namely if a nucleoside derivative could explain the observed mass increment, we first performed a UV cross-linking experiment with and without heavy isotope labeling of U4 snRNA oligonucleotide (nucleotide position 26 -47 (35)) with human spliceosomal protein 15.5K.
Using nonlabeled RNA for UV irradiation, we isolated and sequenced a peptide of the 15.5K protein, 22 LLDLVQQSC-NYK 33, cross-linked to a U nucleotide without and with the mass increment of 152 Da (for data, see XL_U4_ Ex1_naturalrNTP). Fig. 1A and 1B show the MS/MS spectra of the peptide-RNA oligonucleotide cross-link without and with mass increment, respectively. In Fig. 1A the y-and b-type fragment ions clearly reveal the sequence of the cross-linked peptide, and the mass difference between the peptide alone (1422.7177 Da) and the measured precursor mass (1746.7536 Da) corresponds exactly to a uridine nucleotide (324.0359 Da). The spectrum also shows a mass shift in the y-ion series starting from y 4 corresponding to a fragment derived from gas phase dissociation of uridine (ϩC 3 O, ϩ51.9949 Da, see below) identifying the cysteine residue as the cross-linked amino acid. The observed y 5 # to y 8 # peaks correspond to peptide fragment ions that still have the nucleoside attached. Fig. 1B shows the same peptide cross-linked to a U nucleotide, but additionally exhibiting the unexpected mass increment. The observed precursor at m/z 633.9234 3ϩ corresponds to a molecular weight of 1898.7502 Da, a difference of 476.0324 Da from the molecular weight of the peptide alone, which corresponds to a cross-link with [Uϩ152 Da]. Manual interpretation of the spectrum reveals a ϩ378.0556 shift of the y ion series starting with y 4 . The mass shift is consistent with a uridine fragment [ϩ226.0590 Da] plus an additional 152 Da mass increment cross-linked to the cysteine residue (Fig. 1B).
We next performed 15.5K-RNA cross-linking experiments with different 13 C-labeled U4 snRNA oligonucleotides synthesized by in vitro transcription. In each experiment, a different 13 C-rNTP was incorporated, i.e. 13 C-A, 13 C-G, 13 C-U, 13 C-C (for data, see XL_U4_Ex1_C13ATP, XL_U4_Ex1_C13GTP, XL_U4_Ex1_C13UTP, XL_U4_Ex1_C13CTP). If the mass increment was caused by a nucleobase derivative, then the cross-link would show an increased mass shift corresponding to its heavy isotope labeling, which is 4 Da for U and C; and 5 Da for A and G, respectively. None of the experiments employing heavy isotope labeling, however, produced any evidence of a heavy isotope-induced shift in the ϩ152 Da mass increment. In all observed cross-links the observed mass shifts could be readily and conclusively rationalized by 13 C labeling of the nucleobases directly involved in the actual cross-link, without any indication of a heavy isotope shift in the ϩ152 Da mass increment. Fig. 1C shows an example MS/MS spectrum from the cross-linking experiment using 13 C-UTP labeled oligonucleotide (for data, see XL_U4_ Ex1_C13UTP), corresponding to a peptide-RNA (oligo)nucleotide cross-link 22  Evaluation of DTT as the Origin of the Unusual Mass Adducts-According to our second hypothesis, the unexpected mass increment might be caused by the addition or insertion of a small organic molecule, possibly dithiothreitol, which has an elemental composition matching the calculated exact mass difference of 151.9966 Da between regular and modified cross-links. To evaluate this hypothesis we again chose a simple system producing cross-links upon UV irradiation. We selected an SmB/BЈ peptide 9 MLQHIDYRMRCILQDGR 25 from human spliceosomal U small nuclear ribonucleoprotein particles (snRNPs). Sm proteins consist of a highly conserved set of seven proteins (SmB/BЈ, D1, D2, D3, E, F, and G), which are common to all U snRNPs (U1, U2, U4, U5) and assemble as a seven-membered ring around a single stranded U-rich sequence called the Sm site. The selected peptide sequence was observed to cross-link with the Sm site within endogenous U snRNAs during the investigation of protein-RNA interactions in spliceosomal U1 small ribonucleoprotein particles (Fig. 2E, PDB ID: 4PJO, (50,51)). Close inspection of peptide-RNA (oligo)nucleotide cross-links obtained in these studies of endogenous RNPs already showed that a cysteine within the C-terminal half of the peptide ( 19 CILQDGR 25 ) specifically cross-linked to uracil, again with a mass increment of 152 Da (Fig. 2A). We also identified peptide-RNA (oligo)nucleotide cross-links derived from the N-terminal half of the peptide 9 MLQHIDYR 17 but without the 152 Da adduct (Fig.  2B), even in the presence of DTT (for data, see XL_peptide_ Ex2_withDTT). Here, Tyr15 was found as the cross-linked amino acid (Fig. 2B). This result is consistent with previous observations that this mass increment is usually associated with cysteine containing peptide sequences, and strongly points toward a UV-induced cross-linking reaction that includes DTT.
We performed in vitro UV cross-linking with this particular peptide with an Sm site RNA oligonucleotide AAUUUGUGG  (61). A proposed schematic illustration of the peptide-RNA cross-link along with cross-link product of ϩ152 Da is shown on the right. Cysteine as the cross-linked amino acid identified by shift in the y ion series (starting from the y 4 ion), is highlighted by bold letter in the peptide sequence. y-type fragmentation of RNA is referred to as U', w r1, y r1 , y r2 . Positions of 13 C on UTP are marked by orange circles in lower panel. following our standard protocol, both without and with the addition of DTT (for data, see XL_peptide_Ex2_withoutDTT, XL_peptide_Ex2_withDTT). Peptide-RNA (oligo)nucleotide cross-links were observed corresponding to both peptide sequences MLQHIDYR and CILQDGR obtained by tryptic cleavage. Interestingly almost all cross-links derived from the tryptic peptide CILQDGR, but not those of from peptide MLQHIDYR, were observed both with and without the mass increment of 152 Da in the presence of DTT (supplemental Table S1). Without the addition of DTT, none of the side products were observed. Fig. 2C and 2D shows the MS/MS spectra of two peptide-RNA oligonucleotide cross-links obtained from the peptide CILQDGR in the presence and absence of DTT (for data, see XL_peptide_ Ex2_withDTT, XL_peptide_ Ex2_withoutDTT), respectively. The calculated molecular weight difference between the cross-link (1279.4284 Da) and the corresponding peptide (803.3960 Da) is 476.0288 Da. This difference matches to a U nucleotide plus a mass increment of 151.9965 Da. An analysis of the product ion spectrum reveals a complete y ion series. In addition the spectrum contains a and b ion series shifted by a fragment of uracil plus the mass increment of 152 Da including an ion nominally corresponding to b 1 # , identifying the cysteine residue as the site of crosslinking (Fig. 2C). If a peptide's N-terminus is not modified by e.g. acetyl, then b 1 ions are usually not observed under CID conditions as they cannot be stabilized via an oxazolone structure as is the case for higher order b ions (52). We postulate, however, that the large nucleotide moiety stabilizes the charge sufficiently. In the absence of DTT we found a crosslink of 1127.4318 Da corresponding to cross-linked U nucleotide (324.0359 Da) without the mass increment of 152 Da (Fig. 2D).
For further confirmation the UV cross-linking experiment was performed in the presence of perdeuterated DTT (d 10 -DTT). Of the ten deuterium atoms in d 10 -DTT, two belong to the thiol functions and two more to the weakly acidic hydroxyl functions, which are prone to hydrogen exchange in aqueous media. We would therefore expect a mass increment of 158.0342 Da if deuterium labeled DTT was incorporated in cross-links because of the remaining 6 C ␣ -deuteriums. Fig. 3A shows the MS/MS spectrum of a cross-linked peptide-RNA (oligo)nucleotide obtained from the peptide CILQDGR in the presence of DTT (for data, see XL_peptide_ Ex2_withDTT). The calculated molecular weight difference between the cross-link (1937.5335 Da) and the corresponding peptide (803.3960 Da) is 1134.1375 Da. This difference matches the trinucleotide AAU (982.1410 Da) plus a mass increment of 151.9965 Da. An analysis of the fragment spectrum reveals a complete y ion series up to y 6 , again pointing to the cysteine residue as the site of cross-linking. In addition the spectrum contains a number of fragments in the upper m/z region that can be assigned to successive fragmentation of the RNA moiety of the cross-link.
Interestingly the mass increment of 152 Da is present in all RNA-derived fragments down to the smallest units, suggesting that the structure causing the fragment is either close to, or actually part of the site of cross-linking. In the low m/z region the spectrum exhibits marker ions for the adenine nucleobase at m/z 136.0620 and the corresponding nucleotide at m/z 330.0600, respectively (Fig. 3A), corroborating that the cross-link is formed via the uracil nucleobase rather than one of the adenine bases. Uracil itself does not generate strong marker ions but ion series in the high m/z region confirmed that uracil is linked to cysteine either directly or via the 152 Da adduct. Moreover we observed a fragment ion at m/z 265.0314 because of the mass of the uracil base shifted by 151.9963 Da (Fig. 3A). Fig. 3B shows the MS/MS spectrum of the peptide-RNA (oligo)nucleotide cross-link derived from the peptide sequence CILQDGR with the d 10 -DTT-AAU nucleotides (for data, see XL_peptide_Ex2_dDTT). Comparison of the spectrum with that of the nonheavy isotope-labeled species in panel A clearly shows that all fragments incorporating the oligonucleotide moiety are similarly shifted upward by 6 Da owing to the six deuterium atoms in DTT, whereas y-type fragments assigned to the peptide sequence do not exhibit this shift. The results obtained on this model system strongly support the hypothesis that DTT is indeed the origin of the 152 Da mass increment.
Importantly, we observed fragments at m/z 1008.3978 and 1014.4338 ( Fig. 3A and 3B, respectively) corresponding to a 6 Da mass difference in these product ion spectra¸the latter contains deuterated DTT (ϩ6 Da). We assigned these values to the mass of the peptide (803.  (51.9928 Da). We conclude that C 3 O must have derived from internal fragmentation of uracil as indicated in Fig. 4. According to Nelson and McCloskey (53) protonated uracil follows different fragmentation pathways under CID conditions. A neutral loss of HNCO leads to C 3 H 4 NO ϩ (70 Da) and further neutral loss of NH 3 to C 3 HO ϩ with a nominal mass of 53 Da, which consists of the C4, C5, and C6 atoms from uracil. We note that the actual structure of this particular charged product ion is different as the cross-link to the DTT-peptide has to be considered. Therefore we postulate a structure of C 3 H 2 O ϩ covalently linked to the DTT-peptide as depicted in Fig. 4. If we assume a reduction of the double bond between positions 5 and 6 and the DTT-peptide moiety attached to the C5-atom, then a structure of C 3 H 2 O ϩ attached to the nonprotonated DTT-peptide appears highly plausible (Fig. 4). The exact masses of the product ions in Fig. 3A and 3B, with m/z 1008.3978 and 1014.4338, respectively, fit within less than 5 ppm to the predicted structure of the DTT-peptide to uracil and thus strongly support this structure. Note that similar product ions with the same mass increment of 52 Da were observed upon CID fragmentation of other UV-induced peptide-RNA (oligo)nucleotide cross-links containing the crosslinked amino acids glycine, lysine, proline, tyrosine, and valine (U.Z., H.U., unpublished observations) in the absence of DTT (i.e. mass increment of 151.9966 Da), which clearly shows that the product ion mass increment of 52 Da must have been derived from uracil.
DTT-driven Protein-RNA Cross-linking in Intact RNP Complexes-To complement our in vitro experiments on low complexity model systems, we decided to investigate a more complex system to see if additional information could be gained from a more statistical approach evaluating a larger number of peptide-RNA (oligo)nucleotide cross-links. For this purpose a UV cross-linking experiment was performed on a large protein-RNA complex (RNP) from S. cerevisiae obtained by using fused TAP-Cbp20 tag. Because yeast Cbp20 together with Cbp80 binds to capped mRNA, the purified RNP represents a mixture of proteins that have direct or indirect interactions with mRNA. We have recently successfully combined this purification strategy with UV-induced cross-linking and several peptides cross-linked to RNA (22). Of note, in this preparation we used 2 mM DTT in the final buffer used for the purification of the yeast mRNPs, and most of the cysteine containing peptides that were found cross-linked showed the mass increment of 152 Da.
To study the role of DTT in cross-link formation in such a large-scale preparation, we performed TAP-tag purification of the yeast mRNPs according to Kramer et al. (22). The final step of mRNP purification was performed in buffer containing either DTT or TCEP in two parallel experiments (for data, see XL_yeast_Ex3_DTT, XL_yeast_Ex3_TCEP). After UV irradiation of the purified RNPs, proteins were hydrolyzed by trypsin under denaturing conditions and cross-linked peptide-RNA heteroconjugates separated from noncross-linked peptides by size exclusion chromatography. RNA was hydrolyzed to short fragments using RNases A/T1 and benzonase. After desalting by reverse phase chromatography, samples A and B were analyzed by nLC-ESI-MS/MS and data was analyzed using the RNP xl software pipeline (22). Product ion spectra assigned to cross-links were manually evaluated.
In total we identified 99 peptide-RNA (oligo)nucleotide crosslinks containing 59 unique cross-linked peptide sequences (with various combination of cross-linked RNA nucleotides) derived from 43 different proteins (supplemental Table S2 and Table II). From these 59 unique cross-linked peptides, 26 were exclusively found in the preparation with DTT and 10 exclusively with TCEP, respectively. 17 cross-linked peptides were found in both preparations. The majority of peptide-RNA (oligo)nucleotide cross-links derived from the DTT experiment (24 out of 43) exhibit a mass increment of 152. Importantly, 152 Da is also depicted. B, MS/MS spectrum of m/z 972.7957 2ϩ in the presence of d 10 -DTT shows the peptide cross-linked to the trinucleotide AAU with the additional mass of 158 Da. Signals in the high m/z range correspond to shifting of mass of peptide shifted peptide fragment masses by different RNA fragments along with 158 Da mass, and are annotated as in spectrum A. The difference of 6 Da is clearly observed between the two spectra in the high m/z region. C, Proposed structure for peptide-DTT-uracil cross-link formation. Upon UV irradiation at 254 nm, DTT can form covalent bonds with uracil and the thiol group of cysteine.

FIG. 4. Proposed scheme for the cross-linking reaction of uridine/uracil to cysteine/DTT. Panel (A) shows UV
induced peptide-DTT-uridine cross-link formation. Panel (B) shows CID fragmentation of cross-linked uracil according to Nelson and McCloskey (53). Sequential loss of HNCO and NH 3 from uracil result in formation of C 3 H 2 O ϩ covalently linked to DTT-peptide. Abbreviations used in the figure are "P" for peptide and "R" for ribose.
peptide-RNA (oligo)nucleotide cross-links containing the 152 Da mass increment were exclusively found in the presence of DTT in the buffer solutions and all these cross-linked peptides revealed a cysteine as cross-linked amino acid. In three cross-linked peptide RNA (oligo)nucleotides (60S ribosomal protein L35A/B, 40S ribosomal protein S6 A/B, Cruciform DNA recognizing-protein) the cysteine was found to be cross-linked to RNA with and without the 152 mass increment.  Kramer et al. (22) are marked by * sign. Abbreviations used in the table are, TCEP: Tris(carboxyethyl) phosphine-hydrochloride, DTT: dithiothreitol. "-" is denoted for cross-links not identified in the corresponding experiment. "0" is denoted for the nucleotide minus water and "-p" for the loss of metaphosphoric acid (HPO 3 (Fig. 5B). In both cases the fragment spectra are highly similar with almost complete y-ion series. Shifting of the y 5 and y 6 ions by 306 Da and 458 Da, respectively, identified Cys83 as the cross-linked amino acid, which lies within the flexible region of protein.
Mapping of the cross-linked peptide to the crystal structure of S6 -40S ribosomal protein (PDB ID: 3J78, (54)) also reveals the presence of U161 of 18S rRNA in close spatial proximity to Cys83 (Fig. 5C).
For the other peptide-RNA (oligo)nucleotide cross-links it was not always possible to determine the position of the cross-linked amino acid on the peptide sequence accurately owing to the labile nature of the cross-links under CID-MS/MS conditions (22). From 43 identified cross-linked proteins, 32 have been identified in the study of Kramer et al. (22) as well. The majority of the cross-linked proteins derived from ribosomal subunits, 14 are nonribosomal proteins. Among these are cross-linked peptides derived from proteins with a described RNA-binding motif: nucleolar protein 3, polyadenylate-binding protein, single-stranded nucleic acid-binding protein, and Transposon TyH3 polyprotein with cross-linked peptides located in the RNA-binding motif. Further cross-links derived from a nonribosomal protein that does not contain a "classical" RNA binding domain (RBD), i.e. elongation factor 1-alpha and metabolic enzymes including glycerol phosphate phosphohydrolase, phosphogluconate dehydrogenase, pyruvate kinase 1, peroxiredoxin type-2, enolase 1, saccharopine dehydrogenase, and inorganic prophosphatase. Indeed, cross-linking of metabolic enzymes to RNA has been previously described (16,17,22) and the sites we have identified here in enolase 1, pyruvate kinase 1, and inorganic pyrophosphatase were also previously identified by Kramer et al. (22). DISCUSSION We have shown that DTT specifically cross-links cysteine residues within RNA-binding proteins to the uridine nucleobase, and that this reaction requires activation by UV irradiation. DTT acts in a surprisingly efficient manner when compared with "pure" UV-light based photo-cross-linking and thus can be actively used as a protein-RNA cross-linking enhancer in UV light induced cross-linking. This is an entirely novel finding and expands the use of DTT, which is used mainly in the isolation and purification of biochemical preparations or assays. As a strong reducing agent, DTT usually reacts to form a stable six-membered ring with an internal disulfide bond (55). In its reduced state it promotes protein denaturation by disruption of S-S bonds and prevents protein aggregation by preventing or reversing oxidation of cysteine side chains.
We observed that UV-induced cross-links are purely additive in nature, including cysteine cross-links to uracil both without and with DTT. This suggests that the double bond of uracil between positions 5 and 6 is reduced when the free sulfhydryl group (of DTT or cysteine) reacts under UV light with uracil. This finding is in line with earlier and recent detailed investigations of the UV-induced reaction of amino acids within proteins to RNA (22). In previous studies it has been suggested that sulfhydryl groups are photoreactive toward uracil and add preferentially at the 5 position of uracil in its triplet excited state (56,57). This is further corroborated by the observation that UV irradiation of 5-bromouracil in the presence of cysteine or glutathione results in the formation of 5-S-cysteinyl uracil or 5-S-glutathione uracil, respectively (58,59). Addition of DTT at position 6 of uracil is not favored (60).
We therefore have postulated a plausible reaction scheme outlined in Fig. 4 to explain the cross-linking reaction of uridine/uracil to cysteine/DTT. However, our mass spectrometric results do not allow for any conclusions as to how the reaction exactly occurs mechanistically.
In earlier and recent studies we found that the UV-induced protein-RNA cross-linking reaction is not specific to a certain amino acid, as all amino acids except for aspartic acid and glutamic acid have been found to be cross-linkable by UV light to bases of nucleotides ((22), U.Z. and H.U., unpublished results). However, no precise conclusion can be drawn about the cross-linking yield in these studies, owing to a lack of appropriate standards for quantification. Interestingly, when compared with pure UV-induced zero-length cross-links, the UV-induced DTT cross-linking reaction seems to be surprisingly efficient for identification of cross-linked peptides containing cysteine residues. When we compared the identified cross-linked peptide-RNA oligonucleotides derived from the purified yeast pre-mRNA complexes in the presence of DTT or TCEP, we found that 21 peptide-RNA (oligo)nucleotide crosslinks contained cysteine residues with DTT, demonstrating that DTT enhances efficiently the cross-linking of peptides containing cysteine residues to RNA. Even in previous studies where DTT was only present in cell lysis and/or RNP purification buffers, we found a relatively large proportion (19 out of 24) of cross-linked peptides containing cysteine residues that also contained DTT (22).
The DTT-driven cysteine photo-cross-linking of proteins to nucleotides is specific, although the presence of the DTT increment suggests a "spacer-distance" of ca. 10 Å between the C-␣ atom of cysteine and C-5 atom of uracil. Because we show here that cross-linking to nucleobases occurs at the same cysteine with and without DTT adduct in ribosomal proteins (see Table II and Fig. 5) it identifies those cysteine residues that are in close proximity to the RNA. Moreover, in a recent study of the crystal structure of the RNase H domain (without its cognate RNA) of the spliceosomal protein Prp8, two peptides derived from Prp8 were found to be cross-linked to U4 snRNA (46). The apparent cross-linked amino acids Tyr1858 (peptide 1) and Cys1878 (peptide 2) are both found at the base of the same ␤-hairpin loop. The peptide-RNA (oligo) nucleotide cross-links including Cys1878 showed the DTT mass increment.
In summary, we introduce here DTT as a potent protein-RNA cross-linker that makes it possible to link cysteine residues in RNA-binding regions to RNA, and is superior over pure UV light cross-linking of cysteine residues. The DTTmediated protein-RNA cross-linking reveals RNA-binding proteins with their regions that have not been brought to attention in the context of RNA-binding before. Its suitability for similar approaches, e.g. in protein-DNA cross-linking, however, still needs to be evaluated. In addition, a detailed understanding of the reaction mechanism might help overcome the limited reaction yield achieved by allowing for further optimization of the cross-linking conditions.