Automatic Identification Approach for High-Performance Liquid

Jul 19, 2015 - This paper developed a new approach for global FA profiling and identification for HPLC-MRM FA data mining. Mathematical models for ide...
0 downloads 30 Views 1MB Size
Article pubs.acs.org/ac

Automatic Identification Approach for High-Performance Liquid Chromatography-Multiple Reaction Monitoring Fatty Acid Global Profiling Cai Tie,*,‡ Ting Hu,‡ Zhi-Xin Jia, and Jin-Lan Zhang* State Key Laboratory of Bioactive Substance and Function of Natural Medicines, Institute of Materia Medica, Peking Union Medical College and Chinese Academy of Medical Sciences, Beijing 100050, PR China S Supporting Information *

ABSTRACT: Fatty acids (FAs) are a group of lipid molecules that are essential to organisms. As potential biomarkers for different diseases, FAs have attracted increasing attention from both biological researchers and the pharmaceutical industry. A sensitive and accurate method for globally profiling and identifying FAs is required for biomarker discovery. The high selectivity and sensitivity of high-performance liquid chromatography-multiple reaction monitoring (HPLC-MRM) gives it great potential to fulfill the need to identify FAs from complicated matrices. This paper developed a new approach for global FA profiling and identification for HPLC-MRM FA data mining. Mathematical models for identifying FAs were simulated using the isotope-induced retention time (RT) shift (IRS) and peak area ratios between parallel isotope peaks for a series of FA standards. The FA structures were predicated using another model based on the RT and molecular weight. Fully automated FA identification software was coded using the Qt platform based on these mathematical models. Different samples were used to verify the software. A high identification efficiency (greater than 75%) was observed when 96 FA species were identified in plasma. This FAs identification strategy promises to accelerate FA research and applications.

F

docosanoic acid, exhibit the opposite effects. Identifying FAs is thus important for FA analyses. The insufficiency of standards caused by the high FA diversity makes identifying FAs a challenge. Yang et al. developed a method to identify FAs based on FA collision-induced dissociation (CID) fragments.19 This method requires high-quality tandem MS (MS/MS) spectra. The low abundances of endogenous FAs species implies that obtaining high quality MS/MS data remains a challenge that prevents FA identification. Our previous work provided a clue to solve this dilemma. A general MRM condition for any FA was developed using a novel derivation strategy.20 This method removes the requirement for specifically optimized MRM conditions, which may make full FA profile coverage via HPLC-MRM feasible. Combining high sensitivity and full coverage into a FA global profiling method would accelerate research into biomarker discovery and lipidomics. Matrix effects were the most challenging problems facing FA profiling. Even the high selectivity provided by HPLC-MRM inevitably yields complicated mass spectra and FA misidentifications due to the matrix.

atty acids (FAs) are a group of lipid molecules that are critical for regulating many biological processes.1,2 FA disorders induce chaos in organisms. Recent research has connected FAs to many serious diseases, including tumors, diabetes, and heart diseases.3−8 Preliminary studies indicate that FAs serve as potential diagnosis biomarkers that could enhance diagnostic accuracy and facilitate early diagnosis. Both academia and industry require a sensitive, accurate, and reliable FA global profiling strategy to discover and develop further biomarkers. High-performance liquid chromatography−mass spectrometry (HPLC-MS) is the most widely used FA profiling method and has dramatically improved FA analyses.9−13 However, its limited sensitivity is a major challenge for its further application.14 HPLC-multiple reaction monitoring (HPLCMRM) has been developed to achieve higher sensitivity and selectivity.15−18 Using optimized MRM conditions allows HPLC-MRM to serve as a sensitive strategy for endogenous FA analysis. Their high dependence on standards prevents these methods from being applied for global FA profiling. Thus, FA analysis has been unable to achieve both high sensitivity and full coverage. According to recent research, FA bioactivities are strongly connected to their structures. ω-3 and ω-6 FA clusters have been reported to protect the brain and heart. However, saturated FAs, such as lauric acid, myristic acid, and n© XXXX American Chemical Society

Received: February 28, 2015 Accepted: July 19, 2015

A

DOI: 10.1021/acs.analchem.5b00799 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

behenic acid, heptadecanoic acid, myristic acid, and palmic acid) were purchased from Sigma-Aldrich (St. Louis, MO, USA). Ceramide (18:1/10:0) and ceramide (18:1/2:0) were purchased from Avanti Polar Lipids (Alabaster, AL, USA) and introduced as internal standards during pretreatment. Both the “normal” and “heavy” derivatization reagents (2,4-bis(diethylamino)-6-hydrazino-1,3,5-triazine and d20-2,4-bis(diethylamino)-6-hydrazino-1,3,5-triazine, respectively) were kindly provided by Prof. Zhang.21 All solvents were HPLC grade or higher. Methods. FA Analysis. FAs were extracted from plasma via a previously reported protocol.20 Each sample was equally divided between two vials. One was tagged with “normal” reagents, and the other was tagged with “heavy” reagents, following the previously reported method. Parallel samples were mixed equally and analyzed via HPLC-MRM. Ion pairs were set according to the m/z range of interest. Precursor ions were set by increasing each step by 1 Da, and the collected product ions were 254 m/z for “normal”-tagged and 274 m/z for the “heavy”-tagged samples. Data Processing. The RTs and peak areas were extracted using an Agilent MassHunter (Santa Clara, CA, USA). The software was coded with Qt 5.3 (Qt project, Oslo, Norway) on a computer with an Advanced Micro Devices (AMD) Phantom(tm) II X2 560 Processor and 4.00 GB randomaccess memory (RAM). The software was designed to operate under the Windows 7 operating system (OS) using Microsoft Office 2010.

In this work, a fully automatic FA identification approach based on HPLC-MRM data mining that enables sensitive, accurate, and reliable global profiling of FAs was designed. FAs in biosamples were identified by integrating the derivation strategy with multiple filtration methods. These FA structures were further predicted using another model based on the retention time (RT) and molecular weight. Fully automated FA identification software was coded using these models. By eliminating the matrix effects, this software resolved the matrixeffects issue for global HPLC-MRM FA profiling. Such a sensitive, accurate, reliable, and fully covered FA profiling strategy can help reveal the functions of FAs. Further research and applications focusing on FAs are expected. Prior research has found that the RT shifted due to deuterium alternation between parallel peaks of “normal” and “heavy” tagged FA reagents. A filtration technique was developed on the basis of this shift in the RT to identify FAs in biosamples. Scheme 1 shows a multiple-filtration strategy Scheme 1. Flow Chart for Identifying FAs via HPLC-MRM Analysisa



RESULTS AND DISCUSSION Profiles with IRS and Intensity Ratio. The IRS for different FAs were observed to depend on their RT and were relatively independent of the FA structure. Several FAs with different unsaturations and carbon numbers were studied to determine the relationship between the FA IRS and structure. Supporting Information SI.1 shows that the IRS for FAs with different structures could be predicted via a constant linear relationship. This finding indicated that the FA IRSs were mainly determined by the RT, not the saturation degree or conformation. A method for predicting the FA IRS was developed on the basis of this relationship. Any compounds detected via HPLC-MRM were tested for their IRS. Figure 1 shows that 4 peaks were observed for “normal”-tagged ions EIC. The expected IRS (IRSe) for each peak was calculated from the linear relationship forged from the FA standards (Supporting Information SI.2) and observed IRSs (IRSo) obtained from the parallel “heavy” peaks. IRSo was tested with IRSe. Acceptable zones were generated for this test, as shown in Supporting Information SI.2. Peaks 1 and 3 were located in the acceptable zone and used for further tests. Another two peaks (peaks 2 and 4) were denied as FAs because of their disordered IRS. The results of IRS tests for the detected biosample peaks are presented in Table 1. Four 520 m/z peaks were detected with parallel peaks. The IRSo and IRSe for each pair were calculated on the basis of the RTs. Each IRSo was evaluated within the acceptable zone. Only those peaks (pairs 1/2/3) with IRSo peaks in the acceptable zone were regarded as potential FAs and used during the intensity tests. Supporting Information SI.3 shows that the intensities of parallel peaks were similar. Obviously different areas for apparently parallel peaks indicated that they were not parallel. One more parameter, the intensity ratio between parallel peaks, was adapted to improve the accuracy and reliability of the

The data was tested using three filtrations with parameters generated via training experiments. Only those compounds that passed all the three tests were identified as FAs.

a

containing an isotope-induced RT shift (IRS), signal intensities test, and overcounting test used to identify FAs in biosamples. The structures for FAs passing all of these tests were predicted on the basis of their RTs and m/z. This method was validated using biosamples and proved to be an efficient, reliable, and accurate approach for identifying FAs. Global FA profiling with high sensitivity and full coverage is promising due to this work.



EXPERIMENTAL SECTION Materials. N-Hydroxysuccinimide (NHS), 1-ethyl-3-[3dimethylamino-propyl]carbodiimide hydrochloride (EDC), and all of the FA standards (arachidic acid, arachidonic acid, B

DOI: 10.1021/acs.analchem.5b00799 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

an FA out of the three with constant RTs. The others (compounds 1 and 2) exhibited the same FA isotope peaks.

Figure 2. Eliminating overcounting. Peaks with a constant RT were tested. The peaks caused by isotopes and differentially charged ions were determined.

Figure 1. IRS and signal intensity tests. IRS (red), accepted zone (green), and intensity differences (purple) for each peak pair applied during the IRS and signal intensity tests.

Table 2. Over-Counting Identification

a

Table 1. Tests for Compounds with 520 m/z ID

RTn/ min

RTh/ min

IRSo/ min

IRSe/ min

peak arean

peak areah

ratio H/N

1 2 3 4

6.110 6.783 7.500 8.136

5.924 6.578 7.278 null

0.186 0.205 0.222 null

0.182 0.202 0.223 null

1547 224782 65307 null

4884 690189 6732 null

0.317 0.326 0.739 null

acceptable zone

IRSe ± 0.0111

ID

m/z

RT/min

peak area

isotopic states

1 2 3

466 465 464

5.46 5.46 5.46

9597 53 569 142 807

double mono none

Besides the isotopic distribution, replacing protons with other cations, including sodium ions (+23 Da), potassium ions (+39 Da), and ammonium ions (+18 Da), was another source of overcounting. Because of the distinctive mass shifts and constant RT to nonisotopic peaks, these overcountings were identified as being due to isotopic peaks. FA Discovery from Biosamples. We forged a FA identification strategy that integrated the above filtrations. These HPLC-MRM analysis data for the hydrolyzed FA methyl esters and rat plasma were applied to test this strategy. Figure 3A indicates that over 150 peaks were detected in the hydrolyzed FA methyl esters. Treatment via this strategy removed a considerable number of peaks from the matrix. Approximately 40 FAs were found, as shown in Figure 3B. An identification efficiency of at least 73% was achieved. The plasma sample experienced a much more complicated matrix. Figure 3D shows that hundreds of compounds were detected in the plasma using HPLC-MRM. Manually finding FAs from those peaks could take several days. The established strategy identified 96 FAs, as shown in Figure 3E. The detected FAs were mainly distributed from 400 to 650 m/z in the plasma. This result agreed with prior work.14 Some FAs with long chains and branched structures were also detected. Each test efficiency was calculated on the basis of the different samples. Figure 3C,F shows that the IRS and intensity tests removed approximately 70% of the matrix compounds. Approximately 10% of the remaining peaks associated with overcounting were identified and removed during a second step. The strategy provided FA identification from a complicated matrix with high speed, accuracy, and reliability.

0.67−1.5

a

n: normal tagged; h: heavy tagged; o: observed; e: expected. IRSe acceptable zone for IRS and peak areas ratio were calculated as mentioned in Supporting Information SI.2

identification. Figure 1 compares peaks 1 and 3 with their parallel peaks determined on the basis of the intensity. The obviously different response relative to its paired peak removed peak 1 as an FA candidate. Peak 3 was still regarded as an FA candidate. This method was applied to validate the biosamples, as described in Table 1. The intensity ratios for the three peaks which passed the IRS tests were calculated and compared with the acceptable zone. Only pair 3 was within the acceptable zone and thus regarded as a candidate compound and passed on to the following tests. Eliminating Overcounting. A FA could be counted during the above tests multiple times for various reasons. Eliminating this overcounting is important. The natural isotope distribution was one major source for overcounting. Most FAs exhibited approximately 20% monoisotopic peaks that were mainly associated with 13C and 2H relative to the nonisotopic peaks. For highly abundant FAs, double and even triple isotopic peaks were observed. These isotopic peaks were counted as nonisotopic ones, which resulted in overcounting some of the FAs. To avoid this issue, any constant RT peak with an area of less than 40% of the nonisotopic peak was not counted, as shown in Figure 2. Table 2 indicates that only compound 3 was C

DOI: 10.1021/acs.analchem.5b00799 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

Figure 4. FAs identified in the biosamples (A) that conformed to the standards (B). Each FA is shown with the same colors in both chromatographs.

Figure 3. Methyl ester hydrolyzed FA products analyzed via IRS. There were over 150 compounds found by HPLC-MRM. (A) Analysis via IRS resulted in more than 70% of compounds being denied and 40 FAs identified. (B) Plasma samples were also identified. The original data contained more than 300 detected compounds. (D) One third of the IRS with 96 compounds were determined to be FAs. (E) The different test efficiencies were analyzed during both sample identifications. The hydrolyzed FAs formed ester (C) and plasma (F).

this prediction method can be widely applied, especially for samples with low FA abundances. Integrated applications for FA discovery and predication strategies using powerful FA identification methods with complicated matrices were developed.



CONCLUSIONS A fully automatic FA identification strategy that combines FA discovery and structure predication was developed for FA biosample analysis. Using multiple tests guaranteed that the FAs were accurately and reliably discovered. Along with the FA prediction, automatic, efficient, reliable, and accurate FA identification software was achieved. Several HPLC-MRM data mining methods were used to test this approach. FAs were successfully identified with high efficiency, accuracy, and reliability. This approach was believed to surpass the final challenge for HPLC-MRM global FA profiling. A sensitive, accurate, and reliable tool analyzing FA biomarker discovery and lipidomics research will be forged shortly.

FA Structure Prediction. To avoid any dependence on the standards and requirement for high-quality MS/MS spectra, we designed a method to predict FA carbon numbers and degrees of unsaturation based on the RT and m/z ratios. Supporting Information SI.4 shows the simulated FA RTs determined from the degrees of unsaturation and m/z. Twenty-two FAs with different m/z and degrees of unsaturation were used to investigate the relationship. Supporting Information SI.5 indicates that the FA RTs depended on the number of carbon atoms. However, the constant degrees of unsaturation were linearly related to the RTs. The different unsaturation degrees forged a series of parallel linear relationships. These relationships predicted the FA structure. The FAs were distributed around these lines according to their degrees of unsaturation. These relationships yielded both the number of carbon atoms and degree of unsaturation for FAs. Table 3 details the analysis for some of the detected FAs. The prediction results were validated using standards, as shown in Figure 4. These results strongly confirm the predication accuracy and reliability. Without the requirement for high-quality MS/MS spectra,



S Supporting Information *

Contains additional information as noted in the text. The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.analchem.5b00799.



Table 3. FA Structure Predictionsa

a

ASSOCIATED CONTENT

AUTHOR INFORMATION

Corresponding Authors

ID

m/z

RT/min

No.c

n

1 2 3 4 5 6 7 8 9 10 11

538 564 512 540 492 518 520 546 574 576 602

5.598 6.121 5.04 5.373 5.795 6.111 7.500 8.452 8.664 9.402 9.575

20 22 18 20 16 18 18 20 22 22 24

5 6 2 4 0 1 0 0 1 0 1

*Tel.: +86 10 63165235. Fax: +86 10 63017757. E-mail: [email protected]. *Tel.: +86 10 83154880. Fax: +86 10 63017757. E-mail: zhjl@ imm.ac.cn. Author Contributions ‡

C.T. and T.H. contributed equally.

Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS We would like to thank Pro. Zhang for his generous gifts. This work was supported by a grant from the Ministry of Science and Technology of the People’s Republic of China (2014AA021101).

No.c: carbon numbers; n: unsaturation degrees. D

DOI: 10.1021/acs.analchem.5b00799 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry



REFERENCES

(1) Nguyen, C.; Haushalter, R. W.; Lee, D. J.; Markwick, P. R. L.; Bruegger, J.; Caldara-Festin, G.; Finzel, K.; Jackson, D. R.; Ishikawa, F.; O’Dowd, B.; McCammon, J. A.; Opella, S. J.; Tsai, S.-C.; Burkart, M. D. Nature 2014, 505, 427. (2) Liu, S.; Brown, J. D.; Stanya, K. J.; Homan, E.; Leidl, M.; Inouye, K.; Bhargava, P.; Gangl, M. R.; Dai, L.; Hatano, B.; Hotamisligil, G. S.; Saghatelian, A.; Plutzky, J.; Lee, C. H. Nature 2013, 502, 550. (3) Blaak, E. E. Proc. Nutr. Soc. 2003, 62, 753. (4) Munger, J.; Bennett, B. D.; Parikh, A.; Feng, X.-J.; McArdle, J.; Rabitz, H. A.; Shenk, T.; Rabinowitz, J. D. Nat. Biotechnol. 2008, 26, 1179. (5) Lundstrom, S. L.; Balgoma, D.; Wheelock, A. M.; Haeggstrom, J. Z.; Dahlen, S.-E.; Wheelock, C. E. Curr. Pharm. Biotechnol. 2011, 12, 1026. (6) Mancuso, D. J.; Kotzbauer, P.; Wozniak, D. F.; Sims, H. F.; Jenkins, C. M.; Guan, S.; Han, X.; Yang, K.; Sun, G.; Malik, I.; Conyers, S.; Green, K. G.; Schmidt, R. E.; Gross, R. W. J. Biol. Chem. 2009, 284, 35632. (7) Seyfried, T. N.; Shelton, L. M. Nutr. Metab. 2010, 7, 7. (8) Nomura, D. K.; Long, J. Z.; Niessen, S.; Hoover, H. S.; Ng, S.-W.; Cravatt, B. F. Cell 2010, 140, 49. (9) Kamphorst, J. J.; Fan, J.; Lu, W.; White, E.; Rabinowitz, J. D. Anal. Chem. 2011, 83, 9114. (10) Yapa, U.; Prusakiewicz, J. J.; Wrightstone, A. D.; Christine, L. J.; Palandra, J.; Groeber, E.; Wittwer, A. J. Anal. Biochem. 2012, 421, 556. (11) Kita, Y.; Takahashi, T.; Uozumi, N.; Nallan, L.; Gelb, M. H.; Shimizu, T. Biochem. Biophys. Res. Commun. 2005, 330, 898. (12) Buczynski, M. W.; Stephens, D. L.; Bowers-Gentry, R. C.; Grkovich, A.; Deems, R. A.; Dennis, E. A. J. Biol. Chem. 2007, 282, 22834. (13) Dickinson, J.; Murphy, R. J. J. Am. Soc. Mass Spectrom. 2002, 13, 1227. (14) Thomas, M. C.; Dunn, S. R.; Altvater, J.; Dove, S. G.; Nette, G. W. Anal. Chem. 2012, 84, 5976. (15) Pettinella, C.; Lee, S. H.; Cipollone, F.; Blair, I. A. J. Chromatogr. B: Anal. Technol. Biomed. Life Sci. 2007, 850, 168. (16) Gonzalez-Periz, A.; Horrillo, R.; Ferre, N.; Gronert, K.; Dong, B.; Moran-Salvador, E.; Titos, E.; Martinez-Clemente, M.; LopezParra, M.; Arroyo, V.; Claria, J. FASEB J. 2009, 23, 1946. (17) Baker, P. R.; Lin, Y.; Schopfer, F. J.; Woodcock, S. R.; Groeger, A. L.; Batthyany, C.; Sweeney, S.; Long, M. H.; Iles, K. E.; Baker, L. M.; Branchaud, B. P.; Chen, Y. E.; Freeman, B. A. J. Biol. Chem. 2005, 280, 42464. (18) Batthyany, C.; Schopfer, F. J.; Baker, P. R.; Duran, R.; Baker, L. M.; Huang, Y.; Cervenansky, C.; Branchaud, B. P.; Freeman, B. A. J. Biol. Chem. 2006, 281, 20450. (19) Yang, K.; Dilthey, B. G.; Gross, R. W. Anal. Chem. 2013, 85, 9742. (20) Cai, T.; Ting, H.; Xin-Xiang, Z.; Jiang, Z.; Jin-Lan, Z. Analyst 2014, 139, 6154. (21) Tie, C.; Zhang, X.-X. Anal. Methods 2012, 4, 357.

E

DOI: 10.1021/acs.analchem.5b00799 Anal. Chem. XXXX, XXX, XXX−XXX