Distinct energy metabolism of auditory and vestibular PNAS PLUS sensory epithelia revealed by quantitative mass spectrometry using MS2 intensity Kateri J. Spinellia,b,1, John E. Klimekc,1, Phillip A. Wilmarthc, Jung-Bum Shina,b,2, Dongseok Choid, Larry L. Davidc,e, and Peter G. Gillespiea,b,3 aOregon Hearing Research Center, bVollum Institute, cProteomics Shared Resource, dDepartment of Public Health and Preventive Medicine, and eDepartment of Biochemistry and Molecular Biology, Oregon Health and Science University, Portland OR 97239 Edited by David P. Corey, Harvard Medical School, Boston, MA, and accepted by the Editorial Board December 14, 2011 (received for review September 26, 2011) Measuring the abundance of many proteins over a broad dynamic In our experiments, we desired an accurate quantitation method range requires accurate quantitation. We show empirically that, in for relative comparisons. Because varying peptide ionization, MS2 MS experiments, relative quantitation using summed dissociation- selection, and MS2 signal render the relationship between ion-cur- product ion-current intensities is accurate, albeit variable from rent intensity and peptide abundance uncertain, we sought to em- protein to protein, and outperforms spectral counting. By apply- pirically determine the suitability of MS2 intensity for protein ing intensities to quantify proteins in two complex but related quantitation under conditions of high protein complexity. Moreover, tissues, chick auditory and vestibular sensory epithelia, we find to verify our MS2 intensity quantitation methods, we sought a direct that glycolytic enzymes are enriched threefold in auditory epithe- comparison of the molecular composition of two related but distinct lia, whereas enzymes responsible for oxidative phosphorylation biological tissues. Auditory and vestibular inner-ear organs contain are increased at least fourfold in vestibular epithelia. This striking highly specialized sensory epithelia, each consisting of sensory hair difference in relative use of the two ATP-production pathways cells and supporting cells. Although they are morphologically and likely reflects the isolation of the auditory epithelium from its functionally similar, auditory hair cells are stimulated by vibrations NEUROSCIENCE blood supply, necessary to prevent heartbeat-induced mechanical of the whole cochlear sensory epithelium and its underlying basilar disruptions. The global view of protein expression afforded by membrane, whereas vestibular hair cells respond to movement of an label-free quantitation with a wide dynamic range reveals molecular overlying ECM structure. Our results show that MS2 intensity specialization at a tissue or cellular level. reports protein abundance more accurately than spectral counting, and that use of this quantitation method reveals biologically relevant odern MS methods allow parallel detection of thousands of differences in energy use by auditory and vestibular epithelia. Mproteins. With accurate proteome-wide quantitation, per- turbation experiments that test functional interconnections of Results protein networks are feasible (1). Although known proteins can be Label-Free Quantitation with MS2 Intensities. We first sought to accurately quantified by using directed and targeted MS (2), these compare spectral counting to MS2-intensity quantitation under approaches require, respectively, high-resolution instruments or controlled conditions replicating typical experimental situations. substantial experimental groundwork. By contrast, the protein We diluted a mixture of purified proteins into an Escherichia coli “ landscape in many experiments is unknown, necessitating a shot- protein extract (Experimental Procedures), carried out SDS/PAGE ” gun (or discovery) proteomics approach (2), which does not as- and in-gel tryptic digests, then subjected peptides to liquid chro- sume prior knowledge of which proteins vary in concentration or matography/tandem MS (LC-MS/MS) by using an ion-trap mass fi modi cation. spectrometer. Following SEQUEST identification of peptides, Although methods that require cotranslational (3) or post- proteins were identified and assigned summed spectral counts and translational (4, 5) labeling enjoy wide acceptance, not only can MS2 intensities (13). labeling be technically impractical for some experiments, but these c ’ We generated normalized spectral counts ( ) and normalized methods accuracy is often no better than that of label-free meth- MS2 intensities (i) by dividing counts or intensity for each protein by ods (6). Several label-free methods have been introduced for MS the sum of all counts or intensities for all detected proteins (ex- quantitation, including accurate measurement of parent ions with cluding contaminants) for each run. Spectral counts and intensities high-resolution mass spectrometers (7). Even more broadly used is spectral counting (8), whereby the number of times a peptide is identified is considered to be proportional to peptide mass fraction. A drawback for quantitation by spectral counts is that counts saturate for abundant proteins (9). To counter this compression, Author contributions: K.J.S., J.E.K., J.-B.S., L.L.D., and P.G.G. designed research; K.J.S., J.E.K., Shin et al. (10) summed collision-induced dissociation products and J.-B.S. performed research; K.J.S., J.E.K., P.A.W., D.C., and P.G.G. analyzed data; and K.J.S. and P.G.G. wrote the paper. (i.e., MS2 peaks) assigned to each protein. Although two proteins fl of differing abundance, each detected by single peptides, are The authors declare no con ict of interest. measured as equal in concentration by spectral counting, intensity- This article is a PNAS Direct Submission. D.P.C. is a guest editor invited by the Editorial Board. based methods, in principle, could yield greater accuracy if the A fi Data deposition: The microarray data reported in this paper have been deposited in the intensity is proportional to peptide mass (Fig. 1 ). Grif nand Gene Expression Omnibus (GEO) database, www.ncbi.nlm.nih.gov/geo (accession no. colleagues introduced “spectral index” (11), essentially the same GSE32272). measure as proposed by Shin et al., and showed that it outper- 1K.J.S. and J.E.K. contributed equally to this work. formed other label-free methods for quantitation. Colaert et al. 2Present address: Department of Neuroscience, University of Virginia, Charlottesville, showed that spectral index was more accurate, albeit less precise, VA 22908. than spectral counting methods (12). None of these reports dem- 3To whom correspondence should be addressed. E-mail: [email protected]. onstrated, however, whether normalized MS2 intensities exactly This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. report the mass fraction of each protein analyzed. 1073/pnas.1115866109/-/DCSupplemental. www.pnas.org/cgi/doi/10.1073/pnas.1115866109 PNAS Early Edition | 1of10 Downloaded by guest on October 1, 2021 Fig. 1. Quantitation with MS2 intensities. (A) Illustration depicting spectral counts and intensities from two peptides (A1 and B1), each derived from a different protein, one of 10-fold greater abundance. If digest efficiency, chromatography behavior, ionization, and charge are equal, peptide A1 will have 10 times greater MS1 intensity than peptide B1. In MS2 spectra, A1 and B1 are both counted as one spectral count; by contrast, if sampling, collision efficiency, and ion purity are similar, peptide A1 will produce 10 times more total MS2 intensity than peptide B1. CID, collision-induced dissociation. (B and C) Detection of spiked proteins in complex protein mixture. Ten purified proteins (ALB, serum albumin; AMY, α-amylase; CA, carbonic anhydrase; CAT, catalase; LACTA, α-lactalbumin; LGB, β-lactoglobulin; OVAL, ovalbumin; PYG, phosphorylase; TF, transferrin) were added in various mass ratios to an E. coli extract (total protein, 30 μg) and the mixture was subjected to LC-MS/MS. Normalized spectral counts (B) and normalized intensity (C) were calculated for each protein at each dilution. Perfect correspondence between the amount spiked and the normalized detection value gives the unity line (dashed). (D) Relationship between normalized counts and normalized intensity for 1,018 E. coli proteins detected in at least three of 15 runs from the experiment of B and C. Fit is a second-order polynomial (i = 1.3 + 1.8 c + 0.07 c2). for peptides shared by more than one protein were divided among Quantitation of Proteins in Chick Auditory and Vestibular Epithelia. those proteins based on the relative counts or intensities of the un- To apply the intensity-based quantitation method in a cell-bi- shared peptides (14). For each protein, c and i were averaged over ological context, we compared expression of proteins in sensory five biological replicate runs. We plotted log c or log i against the log epithelia of chick auditory (i.e., cochlea) and vestibular (i.e., utricle) of the known mass fraction (wi, the ratio of the mass for each spiked organs at embryonic day (E) 20 to 21, when structural and func- – protein relative to the total protein mass injected; equal to mi divided tional development of both organs is nearly complete (15 18). We by mtot), fitting the data for all proteins linearly (Fig. 1 B and C). chose the chick because of the superb description of its hair-bundle If c or i accurately reflects the mass fraction of the protein of development (19), its sequenced genome (20), and the ability to interest, the slope of the log-log plot will equal 1 and