Validation of cross-links

Cross-link candidates obtained from manual spectra interpretation or database search after precur-sor variants generation were validated in several steps. Validation criteria were refined and expanded in the course of this project and are described in detail in the results section. Important validation criteria are briefly listed below.

Correct assignment of monoisotopic peak and charge state were confirmed by evaluating the survey scan preceding the fragment spectrum under investigation. When data from a non-irradiated control was available, extracted ion chromatograms were compared to confirm that the precursor was not present in the control at significant intensity. Results of an independent Mascot search for peptide identification confirmed that the fragment spectrum did not yield any true positive hit for a noncross-linked peptide. Failure to meet any of the above mentioned criteria led to exclusion of the candidate as a false positive.

The experimental fragment spectrum was compared to predicted fragments of the candidate peptide.

Peptide fragment masses were calculated from the amino acid sequence with ProteinProspector.

In TOPPView, Orbitrap data was directly annotated with search results, experimental signals corresponding to calculated fragments were automatically highlighted. Remaining high intensity signals were manually compared to RNA fragments or peptide–RNA adducts.

Cross-link candidates were rejected when several high intensity signals could not be explained by calculated fragments of the candidate cross-link. Particular emphasis was on peptide fragment series in the higher m/z range, high intensity immonium ions, and RNA marker ions. Cross-linked RNA with two or more nucleotides should yield marker ions of significant intensity, marker ions for A, C and G base were expected to be dominating in the fragment spectrum if they appeared in the cross-linked RNA.

3 Results

UV induced protein–RNA cross-linking and its investigation by mass spectrometry is based on the following key steps:

• isolation or reconstitution of the protein–RNA complex(es)

• UV irradiation

• sample preparation for mass spectrometry (enrichment of cross-linked heteroconjugates)

• analysis by mass spectrometry

• data analysis

While several experimental strategies have been developed for UV cross-linking and mass spectrom-etry, there was further need for optimization and adaptation, especially for more complex biological systems. In addition, while advances in mass spectrometry instrumentation have led to great ad-vances, they have also resulted in a call for adjustments and re-evaluations of existing experimental and data analysis strategies.

In the course of this work, all of the key steps were addressed. Experimental workflows were adjusted and optimized for ribonucleoproteins that had not been previously investigated by UV cross-linking and mass spectrometry. However, the major focus of this work was on data analysis. At the be-ginning of this project, MS data derived from cross-linking experiments was analyzed manually.

MS/MS spectra were assigned by hand, a time-consuming process that requires considerable exper-tise in spectra interpretation. While feasible for small ribonucleoproteins and a limited number of spectra, increasing complexity and MS data amounts called for a new approach. Thus, in parallel with investigations of novel aspects in UV cross-linking and optimization of experimental workflows for several ribonucleoproteins, a data analysis strategy was developed and refined which eventually allowed the identification of cross-linked peptides in searches against entire proteomes.

55

3.1 Cross-linking products of 4-thio-uracil and a novel approach for automated data analysis

One of the major constraints in UV induced protein–RNA cross-linking is the low cross-linking yield. A strategy to increase the cross-linking yield is the use of photo-reactive nucleotides, e.g.

4-thio-uracil, 6-thio-guanine, or halopyrimidines such as 5-bromo-uracil.

In order to identify cross-linked peptide–RNA oligonucleotide heteroconjugates by mass spectrome-try, the mass of the cross-linking product has to be known. For native RNA, cross-linking is mainly additive, i.e., the mass of the cross-linked heteroconjugate is the sum of the peptide and oligonu-cleotide masses (e.g.^[65]). However, it was unknown whether the same is true for RNA substituted with carbonothioyl-containing bases.

We set out to address the two major constraints of cross-linking experiments: The use of a photo-reactive base-analogue, 4-thio-uracil (4SU), was investigated with a focus on cross-linking yield and mass of cross-linking products. In parallel, an approach for the automatization of data analysis was developed. For the intended experiments, a simple test system was needed. The NusB–S10 complex from E. coli was chosen since it had been investigated previously by protein–RNA cross-linking and mass spectrometry in our laboratory^[71]. It plays an important role in transcription antitermination and has an enhanced affinity for BoxA-containing RNA. Co-expression of the protein complex had been established and could be reproduced. More importantly, therrn BoxA-containing oligonucleotide used in the previous study is short and contains several uracils. Therefore, the variant of the same oligonucleotide synthesized with 4-thio-uracils at specific positions could be obtained.

More precisely, a 19mer RNA oligonucleotide containing the corerrnBoxA element (underlined) was cross-linked to the NusB–S10 complex. Cross-linking to the unsubstituted oligonucleotide (upper sequence) had been previously investigated^[71]. We compared these results to cross-linking to the same oligonucleotide in which three uracils in the BoxA element were replaced by 4-thio-uracil (lower sequence).

5’-CAC UGC UCU UUA ACA AUU A-3’

5’-CAC UGC UC(4SU) (4SU)(4SU)A ACA AUU A-3’

3.1.1 Influence of 4-thio-uracil on the cross-linking yield of the NusB–S10-complex The influence of 4-thio-uracil on the cross-linking yield of the NusB–S10 complex was investigated by cross-linking of³²P-labeled oligonucleotides to the protein complex. Two 19mer oligonucleotides, with and without 4SU, were 5’-labeled with [γ-³²P]-ATP and cross-linked to the NusB–S10 complex.

Cross-linking products were separated by SDS-PAGE and visualized by autoradiography (see Figure 3.1). UV irradiation of the proteins in complex with the unsubstituted oligonucleotide (lane 2) at 254 nm led to cross-linking products of both proteins, while no protein bands were observed in the non-irradiated control (lane 1). In contrast, the non-irradiated control of the complex with the 4SU-substituted oligonucleotide already contained cross-linking products (lane 3). This illustrates the high reactivity of 4SU: It cross-links under ambient light, even when protected from light as much as possible during the experiment. Increasing irradiation time at 365 nm (1, 2, 5, and 10 min; lanes

3.1 Cross-linking products of 4-thio-uracil and a novel approach for automated data analysis 57

Figure 3.1: Autoradiography of NusB–S10 cross-linked to ³²P-labeled BoxA containing RNA oligonucleotides with and without 4-thio-uracil. The upper panel shows the autora-diography after 15 min exposure of a Phosphorimager screen, the lower panel shows details of the cross-linking products after 1 h exposure. Lanes 1 and 3 correspond to non-irradiated controls of complexes with unsubstituted and 4SU substituted RNA, re-spectively. Lane 2 shows cross-linking of NusB–S10 to unsubstituted RNA after 10 min irradiation at 254 nm. Lanes 4-7 show cross-linking products of the complex with 4SU-substituted RNA after UV irradiation at 365 nm for the time periods indicated above the gel lanes. Figure originally published in^[104].

4–7) produced higher amounts of cross-linking products. However, a high excess of RNA remains uncross-linked, independent of substitution and irradiation time, and despite the high excess of protein used. This exemplifies the generally low yield of UV induced cross-linking.

The majority of cross-linking products observed after denaturing gel electrophoresis were binary protein–oligonucleotide complexes of either NusB or S10 and the oligonucleotide. Both unsubsti-tuted and 4SU-containing RNA also showed higher-order cross-links. Their exact nature cannot be determined in our experiments.

Detailed investigation on the cross-linking products (lower panel in Figure 3.1) allowed for compar-ison of the cross-linking yields of the complexes with unsubstituted (lane 2) and 4SU-substituted (lane 7) RNA after the same irradiation period. Quantitative analysis of cross-linking product band intensities revealed that the cross-linking yield decreased by about 10% for NusB, while it increased by approximately 50% for S10. Thus, for the S10 protein, 4SU significantly enhances the cross-linking yield. At 254 nm, all nucleotides of the 19mer could undergo cross-linking. In contrast, only the three 4SU nucleotides were excited by irradiation at 365 nm. The slight decrease in the cross-linking yield of NusB could be due to it forming cross-links to nucleotides outside the triple U stretch. Upon substitution and irradiation at higher wavelengths, these cross-links might not form, consequently decreasing the cross-linking yield. However, our experiments clearly illustrate the potential of 4SU to increase the cross-linking yield for some proteins.

3.1.2 Development of a novel approach for automated data analysis

Identification of peptide–RNA oligonucleotide cross-links from mass spectrometry data has been done by manual spectra interpretation (see 2.2.10.4). Interestingly, the majority of fragments in the MS/MS spectra of cross-linked heteroconjugates correspond to the cross-linked peptide. In some cases, intense marker ions of the RNA bases or nucleotides are observed. Rarely, adducts of peptide and RNA or their fragments are detected. Based on these observations, we developed an idea for the identification of cross-linked peptides by database search.