1. Database Search Algorithm for Identification of Intact Cross-Links in Proteins and Peptides Using Tandem Mass Sepctrometry
신성호

2. Abstract
MassMatrix
A new DB search algorithm to identify and characterize intact X-links in proteins and peptides with high confidence
Test with BS3 x-linked Cytochrome C
Five x-links were indentified and verified
Discriminate true positives x-linked PSM from false ones
The distribution of statistical scores for true and false positives
ROC analysis
Search for intact x-links in complex Escherichia coli samples

3. Introduction Indentification of X-links in Proteins
Provide invaluable information regarding a protein’s structure, conformation, and interactions
DB Search Algorithms
False positives need to be controlled
Due to the increased search space for searches with x-links
Traditional DB search program (SEQUEST, Mascot)
Cannot be used for analysis of x-linked proteins/peptides

4. Introduction New DB Search Engine
Identify x-links in proteins and peptides
Include three probability-based scoring algorithms
Provided better sensitivity than Mascot, SEQUEST, OMMSA, X!Tandem for a given specificity
For proteins/peptides without any x-links or disulfide bonds

5. Introduction Validated for peptides and proteins with disulfide bonds
By use of peptide standards with known disulfide bonds and bovine pancreatic ribonuclease A
Tested using data sets
Collected on a LTQ-FT mass spectrometry for the tryptic digests of Cytochrome C x-linked by BS3
Identify and verify five x-links for spatial plausibility by comparison with 3D structure

6. Experimental Section
Material, Sample Preparation, and Mass Spectrometry
Horse heart Cytochrome C, X-linking reagent BS3
The x-linked protein samples were purified by SDS-PAGE
Monomer bands were cut and digested by trypsin
Escherichia coli cells were cultured in LB broth using 200 rpm shaking speed
Nano-LC-MS/MS experiments
On a LTQ-FT mass spectrometry

7. Experimental Section DB Search and Search Parameters
Isotope distributions
Deconvolution to obtain the charge states and monoisotopic m/z values of the precursor ions, during raw data -> mzXML
DB: Cytochrome C protein sequence + decoy sequence randomized Cytochrome C sequences
41 mzXML data files were searched against an Escherichia coli k-12 strain sequence DB containing 4,285 protein sequences

8. Experimental Section The search parameters Enzyme: trypsin
Missed cleavage: 2
Modifications: variable iodoacetamide derivative of cysteine and variable oxidation of methionine
Mass tolerances: 10 ppm for the precursor ions Da for product ions
Maximum number of modification: 2
Peptide length: 5-40 amino acid residues
Score threshold: 5.3 for pp for pptag

9. Experimental Section
Fig 1 shows the structure and reaction of the x-linking reagent
The chemical formula of the x-link between two lysine sites: C8H10O2( Da)
3 dead-end x-links

10. Results and Discussion
Search Algorithm
X-link type
Interchain and Intrachain x-link
Peptides with more than 2 x-links are difficult to characterize
Poor fragmentation and large size
Up to 2 x-links Considered
Peptide type
Type 1: only have interchain x-links
Type 2: only have intrachain x-links
Type 3: both inter- and intrachin x-links
Type 4: circular chins
Dead-end x-links in peptides are considered as modification

11. Results and Discussion
X-link search algorithm
based on the disulfide search algorithm
3 search modes
Exploratory search mode
All occurrences of A and B residues in the protein sequences are considered
Confirmatory search mode
Only the x-links specified in the DB by the user will be considered and searched against experimental data
X-links are coded as “A($i)” and “B($i)”, i is the index number of the specified x-link

12. Results and Discussion
Semiexploratory search mode
A limited exploratory search of x-links will be performed between the amino acid residues labeled($ or $x) in the DB
The process of the MassMatrix
Proteins digestion
In silico based on the specified proteolysis reagents
Fragmentation using the appropriate fragmentation model
CID, ETD fragmentation methods
Each chain undergoes fragmentation independently and internal fragments are not searched
Only product ions created from the rupture of a single bond
When one chain undergoes fragmentation, other(s) will be considered as a modification

13. Results and Discussion
Scoring against the experimental MS2 data
The same as those used for peptides without any x-links and those with disulfide bonds as described previously
3 independent statistical scores, pp, pp2, and pptag
pptag is the best standard
MassMatrix produce protein and peptide match lists
XMapper
Generate x-link assignments from MassMatrix results
Scoring
N: the number of peptides assigned to the x-link
np: the number of spectral matches for peptides p

14. Results and Discussion
Validation of the X-Link Search Algorithm
Data set
A tryptic digest of Cytochrome C x-linked by BS3 on a LTQ-FT
Reagent to protein ratio: 25:1
Final protein concentration: 0.12 mg/mL
MS2: 6,982 spectra
DB (Table 1)
Cytochrome C protein sequence + A reversed Cytochrome C sequences + 20 randomized Cytochrome C sequences
When x-links were considered, theoretical peptides was dramatically increased(2.10 ×106, search time 55s)

15. Results and Discussion
Fig 2 shows 2 representative spectra for x-linked peptides
(a): intrachain, (b): interchain,
*: loss of ammonia, `: loss of water
Fig 3 shows the pptag score distribution for TPs and FPs, and ROC analysis for the PSM identified in MassMatrix
Scoring model can discriminate TPs from FPs
ROC indicate that the algorithm performs well
Area under the curve(AUC): 0.91(with X-links) and 0.92
Good sensitivity and specificity for both types of peptides with and without x-links

16. Results and Discussion
Cytochrome C
Contains 19 lysine residues
Potentially form 171 x-links between 2 lysine
25 x-links assigned
Fig 4 shows the scores of all identified x-links for Cytochrome C are mapped in a heat map using XMapper
A majority of the x-links are background(low occurrence and low scores), light blue and cyan
Background are irrelevant and represent the noise

17. Results and Discussion
X-links formed on the Cytochrome C protein due to its representative 3D structure
Present at higher occurrence and higher abundance
5 nonbackground x-links
5 x-links are further verified by comparison with the 3D structure as shown in Fig 5
The distances of the two lysine residues
K25-K27: 5.3 Å, K86-K87: 3.4Å, K7-K100: 7.7Å
K7-K27: 13.9Å, K99-K100: 13.6Å
Slightly longer than the length of the x-link but reasonable given the differences between crystal and solution-phase structures

18. Results and Discussion
FDR
The background x-links were at a similar level to that of the false positives for decoy proteins
Can be controlled by the target-decoy search strategy
FDR=FPx-link/(TPx-link+FPx-link)
5 nonbackground x-links are survived and all of the background x-links were filter at 5% FDR

19. Results and Discussion
Effect of Different X-Linking Conditions
Different X-linking reagent to protein ratio
1:1, 2.5:1, 5:1, 10:1, 25:1, 50:1, and 100:1
Final protein concentration of 0.12 mg/mL
Fig 6 shows the heat maps of the x-links identified
All 5 nonbackground x-links were identified(≥25:1)
Two or more nonbackground x-links were not identified(<25:1)
Fig 7a shows the dependence of the scores of the five nonbackground x-links
K86-K87 was independent of the ratio
Others scores increased

20. Results and Discussion
Fig 7b shows the number of all the background x-links (FDR 5%)
Total background increases(<5:1)
Becomes much less significant(>5:1)
All background x-links can be filtered at FDR 5%
High x-linking reagent to protein ratio favors the x-link determination in a ratio range of 1:1 to 25:1(0.12mg/mL)
The effect of protein concentration
Reagent to protein ratio: 10:1
Various protein concentrations: 0.06, 0.12, 0.60, 2.4 mg/mL
Fig 8 shows the heat maps of the x-links identified in four sample
Higher protein concentration have higher scores for the 5 nonbackground x-links
Improvement becomes much less significant(>0.60mg/mL)

21. Results and Discussion
Fig 9a shows the dependence of the scores of the nonbackground x-links on the protein concentration
The number of total background x-links increases (≤0.12mg/mL)
The number of background x-links at FDR 5% is independent of the protein concentration and stay low(Fig 9b)
High protein concentration favors the x-link determination experiment and this benefit becomes insignificant (>0.12mg/mL at 10:1)

22. Results and Discussion
X-link Search of A Complex Proteome Sample
The dramatically increased search space, when cross-links are considered
Complex proteome samples for x-links against large protein DB is very challenging
Requires enormous computational resources
Takes significantly longer time
A staged search strategy
Two stages
First: search without considering any x-links
Second: protein matches with significant scores from 1 stage will be searched for x-links

23. Results and Discussion
A staged DB search violates the assumption used in the target-decoy search strategy
Cannot be used to estimate and control false positive rates
In future, nonstaged search will become feasible
Escherichia coli proteome in vitro x-linked by BS3
The samples from two replicate experiments were presepareted by SDS-PAGE
41 bands were cut and in-gel digested with trypsin
Analyzed by LC-MS/MS on a LTQ-FT mass spectrometer
341,613 MS2 spectra

24. Results and Discussion
DB: Escherichia coli K-12 strain protein(4,285 protein)
37,600,000 peptides were calculated (36.70 min)
51,992 PSMs were identified
3,393 were x-linked matches
The peptides without any across-links peptides are dominated
The limited efficiency of x-linking experiment
High complexity of the sample
Further made the identification of x-linked peptides and even more challenging
59 proteins were identified(among 456 proteins) with one or more significant x-links with a score higher than 20

25. Results and Discussion
Table 3 shows the top 20 x-links identified
12 x-links from 9 proteins were verified for spatial plausibility by comparison with the published 3D structures
6 x-links have no available structural data
In summary, only a limited number of x-links can be identified in complex proteome samples using LC-MS/MS
Due to the dominating noncross-linked peptides and high sample complexity
The proteome samples are purified and/or enriched for x-linked peptides

26. Conclusions A new DB search algorithm
Developed to identify intact x-links in proteins and peptides
Based on the validated statistical scoring models
High x-linking reagent to protein ratio favors the x-link determination in ratio of 1:1 to 25:1 at a protein concentration of 0.12 mg/mL
Capable of discriminating true positive from false ones
The distributions of statistical scores and ROC analysis