high-spatial resolution mass - download.xuebalib.comdownload.xuebalib.com/4vh4ffccrjr.pdf ·...

123456789

10111213141516171819202122232425262728293031323334353637383940414243444546474849505152

High-Spatial Resolution MassSpectrometry Imaging: Toward SingleCell Metabolomics in Plant Tissues

Rebecca L. Hansen and Young Jin Lee*[a]

Personal Account

T H EC H E M I C A L

R E C O R D

DOI: 10.1002/tcr.201700027

Chem. Rec. 2018, 18, 65 –77 © 2018 The Chemical Society of Japan & Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Wiley Online Library 65

123456789

10111213141516171819202122232425262728293031323334353637383940414243444546474849505152

Abstract: Mass spectrometry imaging (MSI) is a powerful tool that has advanced ourunderstanding of complex biological processes by enabling unprecedented details of metabolicbiology to be uncovered. Through the use of high-spatial resolution MSI, metabolite localizationscan be obtained with high precision. Here we describe our recent progress to enhance the spatialresolution of matrix-assisted laser desorption/ionization (MALDI) MSI from ~50 mm with thecommercial configuration to ~5 mm. Additionally, we describe our efforts to develop a ‘multiplexMSI’ data acquisition method to allow more chemical information to be obtained on a singletissue in a single instrument run, and the development of new matrices to improve the ionizationefficiency for a variety of small molecule metabolites. In combination, these contributions, alongwith the efforts of others, will bring MSI experiments closer to achieving metabolomic scale.

Keywords: High-resolution imaging, MALDI, Metabolomics, Mass spectrometry, Nanoparticles

1. Introduction

Biological imaging has greatly advanced our understanding ofcomplex biological processes. Several recent Nobel prizes wereawarded to those who advanced imaging technologies tovisualize fine details at cellular and subcellular localizations,such as the 2014 Nobel Prize in Chemistry for super-resolvedfluorescence microscopy. These imaging technologies, how-ever, require the artificial labeling of biological macro-molecules such as green fluorescence protein (GFP),[1] forwhich the 2008 Nobel Prize in Chemistry was awarded,Cy5,[2] or citrine[3] tagging, among others.[4] Chemical orother labeling is not only difficult for small metabolitemolecules, but may also distort the metabolism of the system,and requires the molecules of interest to be identified prior tothe experiment. For example, GFP is useful to monitorknown proteins, but not for other classes of biomolecules.

Mass spectrometry imaging (MSI) has been suggested as auseful alternative approach as it can achieve the chemicalspecificity and sensitivity required for biomolecular imagingexperiments.[5–7] It does not require labeling so any com-pounds present on the tissue can be analyzed, which is insharp contrast with most labeling-based imaging methodswhich necessitate a priori knowledge about well-definedtargets. Additionally, simultaneous imaging of hundreds ofcompounds is possible in MSI, unlike typical optical imagingwhere a maximum of only a few targets can be imaged atonce by using several different fluorescent tag colors.[8,9] MSIcan be applied to visualize the localization of any type ofbiomolecule, including proteins, but it is especially attractivefor small molecule analysis as there are limited alternatives.[10]

In this mini-review, we will introduce our journey indeveloping MSI for plant metabolites, particularly to achievehigh-spatial resolution localization of metabolites at cellular

and subcellular levels. Although any desorption ionizationtechnique can be combined with mass spectrometry toperform MS imaging in microprobe mode (e. g., by rasteringthe sample stage), we adopted matrix-assisted laser desorptionionization (MALDI). Using MALDI, the laser can be focusedto micron size, thus reducing the sampling area (i. e. thepossible spatial resolution) down to subcellular sizes. Whilethe use of an ion beam, as in secondary ion massspectrometry (SIMS), can further reduce the sampling size tothe submicron scale, it fragments molecules into atoms andtherefore is not useful for most biological imaging. StaticSIMS using cluster ion beams has shown to be useful formolecular imaging with spatial resolutions typically about 1–5 microns. Submicron spatial resolution imaging of intactmolecules has been reported using TOF-SIMS,[11] but oftensacrifices sensitivity and mass resolution. Earlier achievementof single-cell level plant imaging by other groups includesmatrix-free LDI by Holscher et al.,[12] and laser ablationelectrospray ionization (LAESI) by Li et al.;[13] however, thespatial resolution was limited in these approaches.

In addition to reducing the laser spot size to achieve highspatial resolution, we have identified two major hurdles incurrent MSI practices. One is the ability to identify unknowncompounds directly on tissues without chromatographicseparation, and the other is the limited sensitivity in smallsampling sizes. We have developed a ’multiplex MS imaging’technique to address the former issue,[14–16] and developedvarious new matrices to address the latter.[17,18] We havedivided this review into three major subjects describing ourefforts: 1) to achieve small laser spot sizes, 2) to developmultiplex MS imaging, and 3) to expand the variety ofcurrently available matrices. The work of others is brieflymentioned only when necessary to better understand ourwork.

[a] R. L. Hansen, Y. J. LeeDepartment of Chemistry, Iowa State University, 35 A Roy J CarverCo–Lab, 1111 WOI Road Ames, IA 50011, United States ofAmericaE-mail: [email protected]

P e r s o n a l A c c o u n t T H E C H E M I C A L R E C O R D


123456789

10111213141516171819202122232425262728293031323334353637383940414243444546474849505152

2. High-Spatial Resolution Mass SpectrometryImaging

Mass spectrometry imaging of biological samples has led tounprecedented information and insight into the spatialdistribution of metabolites for a better overall understandingof physiological processes. Spatial resolution in MALDI MSimaging is determined by several factors, but is largelygoverned by two main variables: 1) the laser spot size and 2)the choice of matrix and its application (i. e. choosing anappropriate matrix for analytes of interest and homogeneitywhen applied to the tissues). To this end there have beenmany efforts to reduce laser spot sizes for MALDI imaging,especially over the last decade. Methods include using apinhole,[19,20] multiple focusing lenses[21] or beam expanders,modifications to beam-delivery optics, coaxial laser illumina-tion,[22] oversampling,[23] optical fibers,[24] and transmissiongeometry setups.[25] For a Gaussian shape laser beam, thediffraction-limited spot size is defined by the followingequation:[20]

Ds ¼ M2 4p

lf

Db

1cosq

where Ds is the diffraction-limited spot size, M2 is the beamquality factor, l represents the laser wavelength, f is the focallength of the focusing lens, Db is the input beam diameter,and q represents the incident angle of the laser beam on thesample. As some instrument parameters cannot be easilychanged for a given commercial instrument (l, q), the otherparameters (M2, f, Db) are often modified to achieve smalllaser spot sizes. It should be noted that the laser spot size isalso affected by the fluence for a Gaussian laser (i. e. a higherfluence results in the desorption of molecules from a largerarea); so, a laser energy slightly above the threshold should beused to minimize the spot size.

2.1. Narrow Diameter Optical Fiber for ~12 mm LaserSpot Size

One early method our group employed to achieve high spatialresolution was via optical fibers.[24] In the linear ion trap massspectrometer with a MALDI source (vMALDI LTQ; ThermoFinnigan), the original setup utilized a 200 mm i.d. opticalfiber to guide the laser beam from the laser source to the ionsource of the mass spectrometer. To achieve higher spatialresolution, the original optical fiber was replaced with a 25mm core multimode optical fiber resulting in a laser spot sizeof ~12 mm, as measured from a burn mark on a thin film ofa-cyano-4-hydroxycinnamic acid (CHCA). One importantconsideration with this method is a rapid decrease in the laseroutput as smaller optical fibers are used, necessitating higherlaser powers to achieve sufficient analyte signals which mayresult in damage to the optical fiber.

2.2. Beam-Delivery Optics Modifications for ~9 mmPractical Laser Spot Size

High mass resolution and MS/MS capability are criticallyimportant to accurately identify compounds. Although somecompounds might show diagnostic MS/MS spectra, high-resolution mass spectrometry (HRMS) is often necessary tocompensate for the lack of chromatographic separation in MSimaging. Additionally, high mass resolution is very usefulwhen searching databases for unknown metabolites. In orderto have these two abilities, in 2009 we purchased a MALDI-ion trap-Orbitrap (Thermo Finnigan) that had just becomecommercialized at the time. This new instrument does notuse optical fibers, but a set of optical lenses, to deliver thelaser beam. The original N2 laser of this instrument does notprovide a Gaussian beam profile, but rather a 50–70 mmelliptical shape which was not appropriate for our intention

Rebecca Hansen received her B.S. fromUniversity of Wisconsin-Platteville in2013. She is currently pursuing her Ph.D.degree in Analytical Chemistry under thesupervision of Dr. Young-Jin Lee at IowaState University. Her Ph.D. research fo-cuses on MALDI-mass spectrometry imag-ing and high-throughput metabolomics

Young Jin Lee received his B.S., M.S., andPh. D. degrees from Seoul NationalUniversity, Korea. After working for LGSemiconductor and Hyundai Electronicsfor three and half years, he did post-

doctoral training at Indiana Universitywith David E. Clemmer. He joined IowaState University in 2008 as AssistantProfessor, and was promoted to AssociateProfessor in 2014. He worked for theproteomics core facility of UC Davis as astaff scientist before joining Iowa State.His current research interests includeMALDI-mass spectrometry imaging, high-resolution mass spectrometry analysis ofbiorenewables, and high-throughput me-tabolomics using MALDI-MS.



123456789

10111213141516171819202122232425262728293031323334353637383940414243444546474849505152

to achieve cellular and subcellular level spatial resolution. Inour initial work, we used a high-power Nd:YAG laser(Quantel; maximum output of 1.4 mJ/pulse) with a deflectorto divert 90 % of the laser beam. Using a 10x beam expander(for a ten times larger Db), we were able to achieve an ~10mm laser spot size,[14] demonstrating the potential for highspatial resolution.

In a later work[26] we used a diode-pumped Nd:YAG laser(Elforlight; maximum output of ~200 mJ/pulse) to bettercontrol the laser energy for MALDI experiments. Here wemade three modifications to reduce the laser spot size basedon the optical design used by the Caprioli group,[20,27] 1)improving the beam shape (M2), 2) expanding the beam priorto focusing (Db), and 3) reducing the focal length (f=125 to100 mm). The laser beam initially has a diameter of~1.2 mm, and after passing through a home-built beamexpander consisting of a focus lens, a pinhole, and acollimating lens, the beam diameter was ~11 mm (Db). Forthis input beam diameter with a 100 mm focal length lens,we were able to achieve a laser spot size as small as ~5 mm.This matches the calculated minimum spot size when M2 = 1,l =355 nm, and q =328. However, at this low laser energywe could not achieve sufficient ion signals and for

experimental purposes we had to increase the laser energy,which resulted in a practical laser spot size of ~9 mm.[26]

Nevertheless, combined with an oversampling method, wecould obtain 5 mm spatial resolution MS images of a maizeleaf with distinct localizations of various metabolites (Fig-ure 1). For example, some flavonoids (luteolin/kaempferoland rutin) are localized in the upper and lower epidermalcells, but maysin is uniquely localized in only the upperepidermal cells, presumably due to its anti-insect herbivoryproperties. In contrast, chloroplast membrane lipids (e. g.,phosphatidylglycerol (PG) and sulfoquinovosyl diacylglycerol(SQDG)) are present only in the photosynthetic mesophyllor bundle sheaths cells. Interestingly, 2-hydroxy-7-methoxy-1,4-benzoxazin-3-one glucoside (HMBOA-Glc) and 2,4-dihydroxy-7-methoxy-1,4-benzoxazin-3-one glucoside (DIM-BOA-Glc) are detected in the mesophyll, as previouslyreported, but only between the vascular bundles. Theselocalizations cannot be detected with a traditional single cellsorting technique followed by metabolomics analysis, suggest-ing the importance of direct in situ analysis.

Figure 1. Optical image and MS images of various metabolites in a maize leaf cross section obtained at 5 mm spatial resolution using a ~9 mm laser spot size andoversampling. Signals are normalized to the TIC at each pixel. Maximum values vary for each image. HexP=hexose phosphate, Hex2 =hexose disaccharide, PG=phosphatidylglycerol, SQDG= sulfoquinovosyl diacylglycerol. Reproduced with permission from ref.[26]



123456789

10111213141516171819202122232425262728293031323334353637383940414243444546474849505152

2.3. Simple, Interchangeable, Multi-Resolution OpticalSystem with 4 mm Laser Spot Size

Most recently our group has further modified our opticssetup to achieve a practical laser spot size of ~4 mm whichenabled 4–5 mm spatial resolution imaging to be accom-plished without the need for oversampling.[16] In the previouswork, a home-built Keplerian optical configuration was usedas a beam expander, but this work uses a commerciallyavailable Galilean configuration beam expander to simplifythe optics. In addition, a 1 mm pinhole was utilized insteadof the 25 mm pinhole due to the difficulties in opticalalignment with such a small pinhole, although the largerpinhole does not remove all the non-Gaussian beamcomponents. Spatial resolutions of 5 mm, 10 mm, or 50 mm,with laser beam sizes of ~4 mm, ~7 mm, or ~45 mm, can beachieved by using a 10x, 5x, or no beam expander,respectively (Figure 2). The commercial beam expanders are

simply inserted or removed from the beam path with alocking screw. This modification results in a substantialbenefit; a simple system to quickly adjust the laser spot sizefor different applications. The final modification in this workwas to reduce the final focal length to 60 mm, the shortestpossible distance from the MALDI plate without modifyingthe mass spectrometer ion optics. The diffraction-limited spotsize in this setup is calculated to be 3.2 mm, and the smallestspot size we achieved was ~3.4 mm, for which there was someion signal. Even when a higher laser energy was used forpractical applications, the laser spot size was increased onlymarginally to ~4 mm.

Figure 2 shows the application of this system to a maizeroot cross-section and compares MS images for three differentspatial resolutions. While analyte localizations can still beseen at 50 mm resolution, it is only once you employ thehigher spatial resolutions (e. g. 10 mm or 5 mm) that finelocalization information is revealed. For example, at 50 mm

Figure 2. a) Optical image of maize root cross-section with morphologic features labeled. b) Optical image of maize root cross-section after all three imaging runs.The 10x beam expander run is shown in yellow, the 5x in red, and the no beam expander in green. c) Combination images of three MS images with different falsecolors. The 5 mm resolution image is enlarged to show fine localization information. The max values used in generating the 5 and 10 mm resolution images are 1 x106, 5 x 105, and 6 x 104 for m/z 184.072, 365.102, and 426.077, respectively, whereas 10 times higher max values are used for 50 mm resolution images. Thescale bar is 100 mm. Reproduced with permission from ref.[16]



123456789

10111213141516171819202122232425262728293031323334353637383940414243444546474849505152

resolution m/z 365.102 (a disaccharide) appears to behomogeneously distributed in the stele, but at the higherspatial resolutions it can be determined that it is localizedwithin the xylem of the stele, as well as the large porousregion of the cortex. This type of localization information canalso be determined for other small metabolites, such as m/z184.072 (phosphocholine) and m/z 426.077 (HDMBOA-Glc).

One limitation of this high-resolution setup is a depth offocus issue that is especially apparent with the 10x beamexpander. Different positions on the MALDI plate as well asvariations in tissue thicknesses result in different laser spotsizes. Although the use of the 5x beam expander results in aslightly lower spatial resolution, the laser spot size is muchmore reproducible across multiple plate positions andvariations in tissue thickness. This depth of focus issue couldbe overcome to a certain degree if an aspheric lens with a60 mm focal length was commercially available. An addi-tional consideration with high spatial resolution imaging isthe reduced sampling volume and thus limited amounts ofanalytes available for analysis.[28] Therefore, as spatialresolution increases, the ability to detect low abundance ionsdecreases, and so for some applications more moderate spatialresolution experiments may be more practical.

The highest spatial resolution we are capable of attaining,3–4 mm, is among the best currently available in MALDI-MSI, and there are only two other groups who have thisability, the Caprioli and Spengler groups. Spengler’s groupachieved high spatial resolution using an atmospheric pressureMALDI source by modifying the laser-focusing optics andusing an aperture to remove peripheral laser intensity.[29] Acapillary was bored through the center of the final focusinglens to coaxially deliver the ions and achieve a final spatialresolution of ~3 mm.[29] The Caprioli group achieved a 5 mmlaser spot size using a similar approach to ours,[27] and ~1 mmlaser spot size with transmission geometry.[30]

3. Multiplex Data Acquisition for MS/MS inMetabolomics Experiments

Traditionally most metabolomics type experiments have beendone using gas chromatography-electron ionization-massspectrometry (GC-EI-MS) or liquid chromatography-tandemmass spectrometry (LC–MS/MS) platforms. These techniquesrequire sample homogenization which loses important spatialinformation. Mass spectrometry imaging can provide thismissing localization information, but lacks the benefit ofchromatographic separation and there is often difficultyconfidently assigning metabolite identities. Tandem massspectrometry (MS/MS or MSn) allows for confident metabo-lite identifications, but in mass spectrometry imaging this

requires multiple tissue sections. This results in increased dataacquisition time and can be problematic when samples arelimited. One solution to this is the use of a multiplex dataacquisition method which allows multiple data acquisitiontypes to be incorporated in a single instrument run.

3.1. Imaging MS Method for More Chemical Informationin Less Acquisition Time

Imaging experiments involving high-resolution mass analyzers(HRMS; e. g. Orbitrap) inherently take more time than time-of-flight (TOF) or other lower resolution mass analyzers (e. g.linear ion trap), and as higher spatial resolution capabilitiesbecome more common, even more analysis time will beneeded. In consideration of this limitation, we developed amultiplex data acquisition method using a hybrid linear iontrap-Orbitrap instrument to both reduce the data acquisitiontime and provide more chemical information. A typical massspectrometry imaging experiment involves collecting a spec-trum of a single scan type at every raster position. A multiplexdata acquisition, however, involves splitting each raster stepinto several spiral steps; each spiral step is then assigned adifferent scan type (e. g. MS, MS/MS, or MSn)[14] (Figure 3a).The various scan types are then repeated in a periodic fashionat each raster step.

In this work, the incorporation of three ion trap (IT)scans after one Orbitrap (FT) scan resulted in a 43 %reduction in analysis time compared to Orbitrap only scans.More importantly, multiplex data acquisition allows MS andMS/MS or MSn scan events to be combined in a singleexperiment. This method allows both metabolite images aswell as structural information to be obtained with one tissuesample. An example of this is shown in Figure 3b for anArabidopsis thaliana flower petal. The FT MS image of m/z447.091 could correspond to either quercetin-3-O-rhamno-side (Q-Rha) or kaempferol-3-O-glucoside (K-Hex), whichcannot be distinguished as they are structural isomers. MS/MS imaging of their distinctive fragment ions (m/z 285 and301) allows these metabolites to be distinguished and revealstheir unique localizations. Furthermore, using MS3 imaging,localization differences of even more complex structuralisomers (e. g. Q-Rha-Rha and K-Rha-Hex) could be deter-mined (Figure 3c).

Data-dependent multiplex MS imaging experiments canalso be performed for the analysis of tissue samples when littleor no information is available about metabolites present. Inthis method the instrument selects ions based on theprecursor MS scan on the fly, and collects MS/MS spectra inthe subsequent spiral steps. With this method it is possible toobtain thousands of MS/MS spectra in a single experiment.Dynamic exclusion, precursor and exclusion mass lists, andother parameters can be optimized to obtain as many MS/MS



123456789

10111213141516171819202122232425262728293031323334353637383940414243444546474849505152

spectra as possible while excluding known contaminants andensuring collection of MS/MS spectra for metabolites ofinterest (if known). Data-dependent multiplex acquisitionallows high-mass resolution MS spectra and many MS/MSspectra to be efficiently collected in a single imaging experi-ment. Although this method does increase the amount ofinformation obtained in a single experiment, it does come ata cost of decreased spatial resolution in the MS scans.

In addition to plant tissues, we have also demonstratedmultiplex MS imaging for the analysis of latent finger-prints.[31] Latent fingerprints are an excellent example oflimited sample availability as oftentimes there is only one or apartial fingerprint available for analysis, and in this casesimultaneous MS and MS/MS acquisition would be impor-tant.

3.2. Multiplex Imaging with Polarity Switching

Mass spectrometric measurements are typically performed ina single polarity, either positive mode or negative mode.However, most analytes are detected with high sensitivity inone mode and significantly suppressed in the other mode.This is due to the chemical nature of the compounds: e. g.,acids are easily deprotonated, thus sensitively detected innegative mode, and bases are easily protonated, thus efficientin positive mode. Therefore comprehensive understanding ofmetabolome or lipidome profiles ideally requires dataacquisition in both positive and negative ion modes.Commercial electrospray ionization (ESI)-MS instrumentshave recently become available with the ability to changepolarity during a run (i. e. positive mode in one second andnegative mode the next second), enabling LC–MS data to beacquired in both modes.

There is no instrument designed for polarity switching inMALDI-MS; however, we were able to incorporate theconcept of polarity switching in MALDI-MSI by extendingthe multiplex data acquisition method to include bothpositive and negative ion modes on a single tissue section.[15]

Typically this would be accomplished by imaging twoseparate tissue samples, one in each polarity. This requirestwice as many samples and sample preparation/data acquis-ition time, as well as the need to correlate features ondifferent images. In this proof of concept experiment eachraster step was split into 9 spiral steps. We acquired a HRMSscan and 4 MS/MS spectra in positive mode followed by apolarity switch to negative mode, and then a HRMS scan and3 MS/MS were acquired in negative mode before the polaritywas switched back to positive mode to start the next rasterstep. The limitation of this approach is the dead time of ~7seconds between each switch; a safety wait-time for theOrbitrap high voltage to turn off before switching to theopposite polarity. Although this was reduced to ~5 seconds

Figure 3. a) Illustration of spiral and raster step plate motions. Separate massspectrometric scan events can be defined for each spiral step and are repeatedat each raster pixel. b) FTMS image (left) of m/z 447.091 which can representeither quercetin-3-O-rhamnoside (Q-Rha) or kaempferol-3-O-glucoside (K-Hex). Corresponding MS/MS images (top right) of the K fragment ion (m/z285) and the Q fragment ion (m/z 301) enable localizations of structuralisomers to be determined. The MS/MS spectrum of the whole petal is shownon the bottom right. Scale bar represents 200 mm. c) Differentiation ofstructural isomers of m/z 593 using MS3 images. Neither the FTMS nor MS/MS images can differentiate the isomers. The MS3 images however show clearlocalization differences for the structural isomers. Reproduced withpermission from ref.[14]



123456789

10111213141516171819202122232425262728293031323334353637383940414243444546474849505152

using a special software patch from the manufacturer, it stillresulted in a total dead time of 10.8 h out of a total dataacquisition time of 18.4 h. The adoption of dual high voltagesupplies, as in commercial ESI-Orbitrap instruments, couldalmost completely remove this dead time. Another minordisadvantage is the systematic mass shift, up to ~25 ppm,that results from data acquisition before the high voltages arecompletely stabilized, however this is easily corrected throughmass calibration with known compounds. Despite somelimitations, we successfully applied this approach to visualizevarious membrane lipids in mouse brain sections, both inpositive and negative mode, as well as acquire MS/MS spectrafor selected lipids in a single imaging experiment.

3.3. Building toward Metabolomic Scale MS Imaging

The molecular coverage of MSI is often limited to a handfulof compounds which is not sufficient to understand theoverall metabolomic changes of a biological system. Thislimitation is a result of two main factors: lack of chromato-graphic separation and matrix-dependent analyte selectivity.To overcome these limitations, we proposed a MSI method-ology that combined multiplex MS imaging data acquisitionwith multiple matrices on consecutive tissue sections toincrease the diversity of chemical compounds that could beimaged and identified.[32] In a simple sense, we used thenatural localization and matrix-dependent selectivity ofanalytes as a way of separation, and therefore utilizedmultiplex MS imaging to perform an LC–MS/MS-likeexperiment. This approach could eventually lead to untar-geted metabolomics analysis using MSI with continuedsuccesses in increasing metabolite coverage.

In this proof of concept experiment a simple four spiralstep pattern was used for each raster step that consisted of aHRMS scan in the low mass range (m/z 50–800), a data-dependent MS/MS scan of the low mass precursor, a HRMSscan for the high mass range (m/z 600–1600), and a data-dependent MS/MS scan for the high mass precursor. Six serialtissue sections of a germinated maize seed, 3 in positive (2,5-dihydroxybenzoic acid (DHB), Fe3O4 nanoparticles (NPs),WO3 NPs) and 3 in negative (1,5-diaminonaphthalene(DAN), 9-aminoacridine (9-AA), Ag NPs), were analyzedwith this data acquisition method. In the end, 104 and 62unique analytes with high-quality MS/MS spectra wereobtained in positive and negative mode, respectively. Addi-tionally, over five hundred unique features were detectedbased solely on accurate mass. We were also able todemonstrate the potential of this technology to better under-stand metabolic biology by visualizing the metabolite local-izations in the glycolysis pathway and tricarboxylic acid cycle(Figure 4). The current limitations of this methodologyinclude 1) the lack of sensitivity for diverse analytes due to

the analyte-selectivity of currently available matrices, 2)difficulties with MS/MS efficiency, and 3) the lack of areliable MS/MS database to improve the throughput andconfidence of metabolite identification.

3.4. Multiplex Data Acquisition without SacrificingSpatial Resolution

One major limitation with multiplex MSI is the loss of spatialresolution as a result of the increased number of spiral steps,which limits its applicability to high spatial resolutionimaging. To this end, we have adapted this data acquisitionmethod to utilize overlapping, rather than spatially separated,spiral steps in order to not sacrifice spatial resolution.[33] It isknown that tissue damage is often minimal with MALDI-MSI, and it has been demonstrated that histochemicalstaining is possible after matrix removal.[34] Hence, as long asthere is enough matrix remaining, multiple MS and/or MS/MS spectral acquisitions should be possible on the same

Figure 4. a) Images of metabolites involved in the glycolysis pathway. G6P=glucose 6-phosphate, F6P= fructose 6-phosphate, F 1,6BP= fructose 1,6-bisphosphate, GADP/DHAP=glyceraldehyde 3-phosphate/dihydroxyaceto-nephosphate, 1,3 BGP=1,3 bisphosphoglycerate, 2PG/3PG=2- and 3-phosphoglycerate, PEP=phosphoenolpyruvate. b) Images of metabolitesinvolved in the TCA cycle. Reproduced with permission from ref.[32]



123456789

10111213141516171819202122232425262728293031323334353637383940414243444546474849505152

sample spot. In this overlapping multiplex MS imagingmethod complete overlap was not possible due to a built-insoftware limitation, so a 1 mm shift between spiral steps wasused.

In a proof-of-concept experiment with standards therewas a gradual signal decrease with multiple data acquisitions,as expected. However, at the 5th spiral step, signal intensitiesremained anywhere between 6–32 % of the signal intensity atspiral step #1, depending on the matrix and analyte. A four-spiral step overlapping multiplex method was then applied tothe 10 mm spatial resolution analysis of a maize leaf crosssection. We successfully demonstrated that high-spatialresolution can be maintained with both high- and low-massrange HRMS, while also acquiring hundreds of high-qualityMS/MS to aid in the identification of visualized metabolites.

4. Development of New Matrices for High-SpatialResolution Imaging

In MALDI-MSI the choice of matrix is critically important tothe types of compounds that can be efficiently desorbed andionized. While the exact effect of matrices on the desorption/ionization process is not well understood, it is known thatdifferent matrices preferentially ionize different classes ofcompounds.[35] For example, CHCA is commonly used forpeptides and small proteins and trihydroxyacetophenone(THAP) is well known for oligosaccharides. Effective matricesshould have good laser absorption, high vacuum stability, noor minimal matrix interference, and be homogeneouslydeposited. Traditional organic matrices such as CHCA,THAP, DAN, and DHB have been most widely used;however, inorganic nanoparticle matrices, such as Ag or AuNPs, TiO2 NPs, and carbon-based NPs, have recently gainedpopularity. For the analysis of low-molecular weight com-pounds (m/z<500) using traditional organic matrices inter-ference from matrix-related ions can be especially significant,and so many efforts have been aimed at the development ofnovel matrices for this low mass range.

4.1. Studies by Others for Small Molecule Matrices

Most current high spatial resolution non-MS approaches tosmall molecule analysis are tailored to analyze a specificmetabolite, not broad classes of compounds as metabolomicstype experiments require. Mass spectrometry imaging has theability to analyze hundreds of compounds in one experiment,but oftentimes the types of molecules that are effectivelyanalyzed are limited by the matrix chosen. Thus, developmentand characterization of efficient matrices that can be used toanalyze a wide variety of metabolites in the low mass range isessential.

Negative mode MS analysis is particularly useful as manysmall molecule metabolites are acidic and can be readilydetected in this mode. One important organic matrix fornegative mode analysis of small molecules is 9-AA. 9-AA is amoderately strong base and it readily abstracts a labile protonfrom the small molecule producing the deprotonated species,[M�H]�. This matrix was successfully demonstrated for lowmolecular weight compounds with acidic protons such aspesticides and benzodiazepines.[36] It has also used in yeastsingle cell analysis to effectively ionize species such as ADPand acetyl-CoA, which are important energy-related com-pounds[10]. Another basic matrix suggested for negative modeanalysis was 1,8-bis(dimethyl-amino)naphthalene (DMAN).DMAN produced no significant background peaks and wasdemonstrated to analyze fatty acids, amino acids, vitamins,and plant hormones.[37] Despite its success in ionizing thesetypes of compounds, it was found to be extremely volatilewhich lead to significant contamination and thus was notsuitable for longer measurements, such as in MS imaging.[38]

A new strong base matrix, 1,8-di(piperidinyl)-naphthalene(DPN), has been recently developed to avoid this issue.[39]

Nanoparticles have also been used as matrices for smallmolecule analysis in LDI-MS, or nanoparticle-assisted laserdesorption/ionization (NALDI) mass spectrometry.[40] NPmatrices, in contrast to organic matrices, generally do nothave significant background in the low mass range whichmakes them very useful for small molecule analyses. Forexample, Shrivas et al. reported the use of TiO2 nanoparticlesfor imaging low molecular weight metabolites (LMWM) inmouse brain tissues.[41] The TiO2 NPs enabled the detectionof more unique LMWMs than with DHB (179 vs 4). SilverNPs have been successfully used for the visualization of fattyacids in mouse liver and retinal samples, which were notdetected using a traditional organic matrix, DHB.[42] Re-cently, graphene oxide was demonstrated for imagingLMWM and lipids in negative ion mode on mouse braintissue.[43] In this work, 212 out of 596 potential smallmolecule metabolites, including lipids, were successfullyidentified through database searches. Nanopore surfaces, suchas desorption/ionization on silicon (DIOS),[44] nanostructure-initiator mass spectrometry (NIMS),[45] and silicon nanopostarray (NAPA),[46] can also be used for small moleculesanalysis. These surfaces also do not display any matrixbackground, but they have a significant limitation for tissueanalyses; the laser beam must penetrate the tissue before itcan be absorbed by the underlying nanopore surface. Thisgenerally requires the tissue to be very thin (< 5 mm),although NAPA was successfully demonstrated for 10 mmthick tissue imaging.[47] Bionanostructured surfaces, such asthose from microalgae cell walls or commercialized celite,have also been used as matrices in LDI-MS.[48]



123456789

10111213141516171819202122232425262728293031323334353637383940414243444546474849505152

4.2. 1,5-Diaminonaphthalene (DAN) for MALDI-MS ofSmall Molecules

DAN had previously been reported as an effective matrix forphospholipids,[49] as well as other higher molecular weightcompounds, but had not been reported for low-molecularweight (LMW) analytes (MW<600 Da). As DAN has ahigh basicity and minimal matrix peaks, we tested DAN forthe analysis of LMW compounds in negative mode, andcompared its effectiveness to 9-AA, CHCA, and DHB.[18] Inthe application to LMW standards, CHCA and DHB spectraare largely dominated by matrix peaks and 9-AA inducessignificant fragmentation of phosphorylated nucleotides.DAN displayed higher signal intensities for small acids,phosphate containing compounds, and amino acids, although9-AA showed higher efficiency for UDP-glucose. For lipidstandards both DAN and 9-AA were efficient; however, 9-AAhad significant signal suppression for some phospholipids.Finally, when applied to maize leaf cross-sections, DAN wassuccessfully able to visualize a variety of metabolites includingamino acids, small organic acids, flavonoids, benzoxazinoids,and sulfolipids. These metabolites were imaged with distinctlocalizations at a generally higher signal intensity than with 9-AA, thus demonstrating the utility of DAN as a MALDImatrix for small metabolites. Now DAN has become one ofour matrices of choice for negative mode analysis.

4.3. Large Scale Nanoparticle Matrix Screening for SmallMolecule Analysis

The use of nanoparticles as matrices for LDI-MS analysisdates back to Tanaka’s work with cobalt powder for proteinanalysis,[50] but NPs have gained renewed attention in recentyears with the increased availability of various syntheticroutes. NPs are vacuum stable, have good UV laserabsorption, can be homogeneously applied, and generallyhave little to no matrix background in the low mass region.Despite these advantages, most studies have focused on theapplication of one or two NPs to a limited number ofcompounds, and there had been no large-scale, systematicstudy comparing many NPs for a range of small molecules.

In order to gain a better understanding of the overalltrends in the analyte-selectivity of NPs, we performed a large-scale screening of NPs for MALDI-MS applications. In thiswork, thirteen NPs, including metal oxides, carbon-based,and metal NPs, were compared for the analysis of two dozensmall molecule metabolites, including small organic acids,phosphate-containing compounds, sugars, amino acids, avitamin, a terpene, a fatty acid, and phospholipids[17] (Fig-ure 5). Select organic matrices were also included forcomparison: DHB and DAN in positive mode and DAN and9-AA in negative mode. In positive ion mode, nanoparticles

generally outperform organic matrices. In negative mode,DAN is more efficient for most analytes, although some NPsshow comparable signals. We anticipate that these screeningresults, which indicate the effectiveness of the NPs in theanalysis of different classes of compounds, will be useful inthe selection of matrices for particular applications.

We proposed a thermal desorption model modified fromSchurenberg et al.[51] that partially explains the nanoparticle-assisted laser desorption ionization (NALDI) efficiencies ofmetal oxide and diamond NPs. Briefly, these NPs have goodlaser absorption, high heat capacity, and low thermalconductivity which enables them to be heated to hightemperatures and thus allows efficient desorption of nearbyanalytes. Some trends in Figure 5 can be qualitativelyexplained by this model. For example, WO3 NPs have thelowest thermal conductivity among the studied NPs, resultingin the highest temperatures and the most observed fragmenta-tion of some metabolite standards. Some observations, suchas analyte-dependent specificities, however, cannot be ex-plained with this model suggesting that other factors, such assurface adsorption also play a role.

Capping with organic compounds is commonly used formetal NPs due to their tendency to aggregate; however,capping can often result in significant interference and ionsuppression in small molecule analysis.[52] In this study, weused non-capped metal NPs suspensions that were preparedimmediately before spray-deposition. This resulted in partialsuccess, although some aggregation was apparent and causedsignal loss, especially as the time between suspensionpreparation and spray deposition increased. An alternative toNP suspensions is the use of physical vapor deposition (PVD)(or sputter coating) to create NPs in situ, which has beendemonstrated for Pt,[53,54] Au,[55,56] and Ag.[57–59] We arecurrently working to compare several metal NPs produced viaPVD for their efficacy in small molecule analysis.

4.4. Binary Organic-Inorganic Matrix Mixture

Due to the matrix selectivity of MALDI analyses, it is difficultto obtain complete metabolite profiles with a single analysis.One way to visualize a wider range of metabolites is to usemultiple matrices on serial tissue sections as we havepreviously discussed[32]. This approach, while useful, increasesanalysis time and sample consumption. Another method is touse combinations of matrices, so-called binary matrices, suchas a mixture of DHB and CHCA[60] or Fe3O4 NPsfunctionalized with CHCA or DHB,[61] among others.

The suppression of TAGs by PCs is a well-known issue inMALDI analyses. Currently, DHB is the most commonlyused matrix for the analysis of PCs, but is not an effectivematrix for TAGs. In our nanoparticle screening, we foundFe3O4 NPs were a good matrix for TAGs, but PCs were



123456789

10111213141516171819202122232425262728293031323334353637383940414243444546474849505152

significantly suppressed. This issue is a serious limitation forany study that requires the simultaneous analysis of TAGsand PCs, such as TAG synthesis in plants where PCs are anintermediate in the pathway. To overcome this limitation wehave developed a simple binary matrix mixture consisting ofnon-functionalized Fe3O4 NPs and DHB that enables theefficient detection of both species in the same sample.[62] Thebinary matrix was compared to DHB or Fe3O4 NPs only firstwith a PC and TAG standard mixture, then for TAG and PCanalysis in a germinated maize seed. As expected, DHBefficiently ionized PCs but not TAGs, Fe3O4 NPs wereefficient for TAGs but not PCs, and the binary matrix couldeffectively ionize both lipid species. Additionally, phosphati-dylethanolamine (PE), phosphatidylinositol (PI), and digalac-tosyldiacylglycerol (DGDG) lipid species that are efficientlyionized by Fe3O4 NPs were also efficiently detected with thebinary matrix. Particularly interesting, large polysaccharidespecies were much more efficiently detected with the binarymatrix than with either of the two matrices individually,suggesting a synergistic effect of the binary matrix. When amultiplex MS imaging approach was adopted, the binarymatrix was able to produce more high quality MS/MS spectra(5026 from 79 unique compounds) compared to DHB(4458 from 60 unique compounds) or Fe3O4 (2915 from 64

unique compounds) further emphasizing the utility of thisbinary matrix.

4.5. Matrix Recrystallization of Lipids for High-SpatialResolution

A critical consideration of high spatial resolution MSI is thelimited amount of analytes present in a small sampling area.Thus, it is important to develop sample preparation methodsto enhance ion signals without analyte delocalization.Sublimation of organic matrices produces very small crystalsizes, but can display poor sensitivity due to the lack ofanalyte incorporation into the crystals.[63,64] It has beendemonstrated that a recrystallization step can enhance thesensitivity of proteins without inducing delocalization ofanalytes,[65] but this has been demonstrated only for animaltissues. We successfully applied this recrystallization strategyto plant tissues with DAN as a matrix and demonstrated nodelocalization was present at 10 mm spatial resolution.[66]

Recrystallization with 5 % isopropyl alcohol at 55 8C for 2minutes was most optimal and increased ion signals by threetimes when applied to the imaging of maize leaves. Moreimportantly, we found methanol was not a good recrystalliza-tion solvent due to side reactions with phosphatidic acids,

Figure 5. Summary of nanoparticle screening for small molecule metabolite analysis. Ion signals are normalized to the highest ion signal for each analyte andshown as a heat map. WO3 NPs have significant matrix background in negative mode and were not used for the final screening. An asterisk indicates a fragmention with the precursor shown in parenthesis. Pcho=phosphocholine, PEP=phosphoenolpyruvic acid, G3P=glycerol 3-phosphate, G6P=glucose-6-phosphate,PA=phosphatidic acid, PC=phosphatidylcholine, TAG= triacylglycerol, DAG=diacylglycerol, PE=phosphatidylethanolamine, PG=phosphatidylglycerol, andPI=phosphatidylinositol. Reproduced with permission from ref.[17]



123456789

10111213141516171819202122232425262728293031323334353637383940414243444546474849505152

indicating some care must be taken when using recrystalliza-tion methods.

5. Summary and Outlook

We have successfully achieved five micron high-spatialresolution MS imaging using a four micron laser spot sizeand demonstrated the utility of this optical setup for imagingmaize root cross sections. We anticipate three micron MSimaging could be possible if further optics modifications weremade to obtain a 2–2.5 mm laser spot size. Robustidentification of metabolites will be a primary challenge formass spectrometry imaging moving forward, particularly inplants as it is anticipated that over 200,000 metabolites arepresent in the plant kingdom, with the majority of themcurrently unknown. Accurate mass information combinedwith MS/MS data will be able to partially overcome thislimitation. The multiplex imaging approach that we havedeveloped and demonstrated for a variety of applications willbe particularly useful to obtain this information. Along withMS imaging, in-parallel metabolomics analysis using tradi-tional approaches (i. e. GC-MS or LC–MS/MS) might benecessary for the confident identification of some of thesecompounds. Additionally, the continuous development ofnew matrices will be necessary as the greatest limitation insensitivity comes from deficiencies in ionization efficiency.

When we started this project in 2008 we were advised notto use the term ’metabolomics’, which was thought to beimpossible in MS imaging. After more than a decade of the’-omics’ boom, we all recognize that it is impossible toachieve a ’complete collection’ of all small molecules orproteins (i. e. ‘true’ metabolomics or proteomics). Therefore,metabolomics now often simply represents a ’large scaleanalysis of metabolites’. As we demonstrated in our recentwork, MS imaging in the metabolomics scale is not very faraway and could be accomplished in the near future. Oncethis is achieved, MS imaging in the metabolomics scale willhave the added benefit of allowing differentiation ofmetabolites at the cellular or subcellular level through high-spatial resolution imaging. Some serious bottlenecks need tobe overcome, however, to achieve metabolomics scale MSI,which include robust identification of unknown metabolitesand further development of new matrices to allow bettervisualization of more classes of compounds. In terms ofmetabolite identification, recently there has been a surge intool developments such as MetFrag, iMet, CFM-ID, andMS-FINDER, among others, that are searchable based on insilico MS/MS fragmentation which will prove useful forconfident metabolite identifications.[67–70] Identification ofunknown metabolites, however, still remains a daunting task,but recent efforts to standardize reporting in metaboliteannotation is a crucial milestone to this end.[71] Along with

various innovations in genomics technologies, high-spatialresolution MSI will revolutionize our understanding of thecooperative and antagonistic effects among metabolites,which are programmed by the genetics of multicellularorganisms, by revealing unprecedented details of metabolicbiology.

Acknowledgements

This work is partially supported by the College of LiberalArts and Science, Iowa State University.

References

[1] O. Shimomura, F. H. Johnson, Y. Saiga, J. Cell. Comp. Physiol.1962, 59, 223.

[2] M. J. Rust, M. Bates, X. Zhuang, Nat. Methods 2006, 3, 793.[3] B. Hein, K. I. Willig, S. W. Hell, Proc. Natl. Acad. Sci. U. S. A.

2008, 105, 14271.[4] M. Fernandez-Suarez, A. Y. Ting, Nat. Rev. Mol. Cell Biol.

2008, 9, 929.[5] T. C. Baker, J. Han, C. H. Borchers, Curr. Opin. Biotechnol.

2017, 43, 62.[6] D. Sturtevant, Y.-J. Lee, K. D. Chapman, Curr. Opin.

Biotechnol. 2016, 37, 53.[7] B. A. Boughton, D. Thinagaran, D. Sarabia, A. Bacic, U.

Roessner, Phytochem. Rev. 2016, 15, 445.[8] M. Bossi, J. Folling, V. N. Belov, V. P. Boyarskiy, R. Medda, A.

Egner, C. Eggeling, A. Schonle, S. W. Hell, Nano Lett. 2008,8, 2463.

[9] W. M. Bates, B. Huang, G. T. Dempsey, X. Zhuang, Science2007, 317, 1749.

[10] A. Amantonico, J. Y. Oh, J. Sobek, M. Heinemann, R. Zenobi,Angewandte Chemie-International Edition 2008, 47, 5382.

[11] Q. P. Vanbellingen, A. Castellanos, M. Rodriguez-Silva, I.Paudel, J. W. Chambers, F. A. Fernandez-Lima, J. Am. Soc.Mass Spectrom. 2016, 27, 2033.

[12] D. Hoelscher, R. Shroff, K. Knop, M. Gottschaldt, A.Crecelius, B. Schneider, D. G. Heckel, U. S. Schubert, A.Svatos, Plant J. 2009, 60, 907.

[13] Y. Li, B. Shrestha, A. Vertes, Anal. Chem. 2008, 80, 407.[14] D. C. Perdian, Y. J. Lee, Anal. Chem. 2010, 82, 9393.[15] A. R. Korte, Y. J. Lee, J. Am. Soc. Mass Spectrom. 2013, 24,

949.[16] A. D. Feenstra, M. E. Duenas, Y. J. Lee, J. Am. Soc. Mass

Spectrom. 2017.[17] G. B. Yagnik, R. L. Hansen, A. R. Korte, M. D. Reichert, J.

Vela, Y. J. Lee, Anal. Chem. 2016, 88, 8926.[18] A. R. Korte, Y. J. Lee, J. Mass Spectrom. 2014, 49, 737.[19] R. M. Caprioli, T. B. Farmer, J. Gile, Anal. Chem. 1997, 69,

4751.[20] A. Zavalin, J. Yang, R. Caprioli, J. Am. Soc. Mass Spectrom.

2013, 24, 1153.[21] B. Spengler, M. Hubert, J. Am. Soc. Mass Spectrom. 2002, 13,

735.



123456789

10111213141516171819202122232425262728293031323334353637383940414243444546474849505152

[22] P. Chaurand, K. E. Schriver, R. M. Caprioli, J. Mass Spectrom.2007, 42, 476.

[23] J. C. Jurchen, S. S. Rubakhin, J. V. Sweedler, J. Am. Soc. MassSpectrom. 2005, 16, 1654.

[24] J. H. Jun, Z. Song, Z. Liu, B. J. Nikolau, E. S. Yeung, Y. J.Lee, Anal. Chem. 2010, 82, 3255.

[25] A. Zavalin, E. M. Todd, P. D. Rawhouser, J. Yang, J. L. Norris,R. M. Caprioli, Journal of mass spectrometry : JMS 2012, 47, i.

[26] A. Korte, M. Yandeau-Nelson, B. Nikolau, Y. Lee, Anal.Bioanal. Chem. 2015, 407, 2301.

[27] A. Zavalin, J. Yang, A. Haase, A. Holle, R. Caprioli, J. Am.Soc. Mass Spectrom. 2014, 25, 1079.

[28] M. Wiegelmann, K. Dreisewerd, J. Soltwisch, J. Am. Soc. MassSpectrom. 2016, 1.

[29] M. Koestler, D. Kirsch, A. Hester, A. Leisner, S. Guenther, B.Spengler, Rapid Commun. Mass Spectrom. 2008, 22, 3275.

[30] A. Zavalin, J. Yang, K. Hayden, M. Vestal, R. M. Caprioli,Anal. Bioanal. Chem. 2015, 407, 2337.

[31] G. B. Yagnik, A. R. Korte, Y. J. Lee, J. Mass Spectrom. 2013,48, 100.

[32] A. D. Feenstra, R. L. Hansen, Y. J. Lee, Analyst 2015, 140,7293.

[33] R. L. Rebecca, Y. J. Lee, J. Am. Soc. Mass Spectrom. 2017,doi:10.1007/s13361-017-1699-7.

[34] F. Deutskens, J. Yang, R. M. Caprioli, J. Mass Spectrom. 2011,46, 568.

[35] M. Gluckmann, A. Pfenninger, R. Kruger, M. Thierolf, M.Karasa, V. Horneffer, F. Hillenkamp, K. Strupat, Int. J. Massspectrom. 2001, 210–211, 121.

[36] R. L. Vermillion-Salsbury, D. M. Hercules, Rapid Commun.Mass Spectrom. 2002, 16, 1575.

[37] R. Shroff, A. Svatos, Anal. Chem. 2009, 81, 7954.[38] M. Eibisch, R. Suß, J. Schiller, Rapid Commun. Mass Spectrom.

2012, 26, 1573.[39] Wei, A. Svatos, RSC Advances 2016, 6, 75073.[40] C. Y. Shi, C. H. Deng, Analyst 2016, 141, 2816.[41] K. Shrivas, T. Hayasaka, Y. Sugiura, M. Setou, Anal. Chem.

2011, 83, 7283.[42] T. Hayasaka, N. Goto-Inoue, N. Zaima, K. Shrivas, Y.

Kashiwagi, M. Yamamoto, M. Nakamoto, M. Setou, J. Am.Soc. Mass Spectrom. 2010, 21, 1446.

[43] D. Zhou, S. Guo, M. Zhang, Y. Liu, T. Chen, Z. Li, Anal.Chim. Acta 2017, 962, 52.

[44] E. P. Go, J. E. Prenni, J. Wei, A. Jones, S. C. Hall, H. E.Witkowska, Z. X. Shen, G. Siuzdak, Anal. Chem. 2003, 75,2504.

[45] T. R. Northen, O. Yanes, M. T. Northen, D. Marrinucci, W.Uritboonthai, J. Apon, S. L. Golledge, A. Nordstrom, G.Siuzdak, Nature 2007, 449, 1033.

[46] B. N. Walker, J. A. Stolee, A. Vertes, Anal. Chem. 2012, 84,7756.

[47] S. A. Stopka, C. Rong, A. R. Korte, S. Yadavilli, J. Nazarian,T. T. Razunguzwa, N. J. Morris, A. Vertes, Angew. Chem. Int.Ed. 2016, 55, 4482.

[48] T. Jaschinski, A. Svatos, G. Pohnert, Rapid Commun. MassSpectrom. 2013, 27, 109.

[49] A. Thomas, J. L. Charbonneau, E. Fournaise, P. Chaurand,Anal. Chem. 2012, 84, 2048.

[50] K. Tanaka, H. Waki, Y. Ido, S. Akita, Y. Yoshida, T. Yoshida,T. Matsuo, Rapid Commun. Mass Spectrom. 1988, 2, 151.

[51] M. Schurenberg, K. Dreisewerd, F. Hillenkamp, Anal. Chem.1999, 71, 221.

[52] Z.-J. Zhu, R. Tang, Y.-C. Yeh, O. R. Miranda, V. M. Rotello,R. W. Vachet, Anal. Chem. 2012, 84, 4321.

[53] H. Kawasaki, T. Ozawa, H. Hisatomi, R. Arakawa, RapidCommun. Mass Spectrom. 2012, 26, 1849.

[54] T. Ozawa, I. Osaka, S. Hamada, T. Murakami, A. Miyazato,H. Kawasaki, R. Arakawa, Anal. Sci. 2016, 32, 587.

[55] M. Dufresne, J.-F. Masson, P. Chaurand, Anal. Chem. 2016,88, 6018.

[56] C. Elsner, B. Abel, Advances in Chemical Engineering andScience 2016, 6, 584.

[57] L. Xu, M. Kliman, J. G. Forsythe, Z. Korade, A. B. Hmelo,N. A. Porter, J. A. McLean, J. Am. Soc. Mass Spectrom. 2015,26, 924.

[58] N. Lauzon, M. Dufresne, V. Chauhan, P. Chaurand, J. Am.Soc. Mass Spectrom. 2015, 26, 878.

[59] S. N. Jackson, K. Baldwin, L. Muller, V. M. Womack, J. A.Schultz, C. Balaban, A. S. Woods, Anal. Bioanal. Chem. 2014,406, 1377.

[60] S. R. Shanta, L.-H. Zhou, Y. S. Park, Y. H. Kim, Y. Kim, K. P.Kim, Anal. Chem. 2011, 83, 1252.

[61] P.-C. Lin, M.-C. Tseng, A.-K. Su, Y.-J. Chen, C.-C. Lin, Anal.Chem. 2007, 79, 3401.

[62] A. D. Feenstra, K. C. O’Neill, G. B. Yagnik, Y. J. Lee, RSCAdvances 2016, 6, 99260.

[63] J. A. Hankin, R. M. Barkley, R. C. Murphy, J. Am. Soc. MassSpectrom. 2007, 18, 1646.

[64] W. Bouschen, O. Schulz, D. Eikel, B. Spengler, RapidCommun. Mass Spectrom. 2010, 24, 355.

[65] J. Yang, R. M. Caprioli, Anal. Chem. 2011, 83, 5728.[66] M. E. Duenas, L. Carlucci, Y. J. Lee, J. Am. Soc. Mass Spectrom.

2016, 27, 1575.[67] S. Wolf, S. Schmidt, M. Muller-Hannemann, S. Neumann,

BMC Bioinformatics 2010, 11, 148.[68] A. Aguilar-Mogas, M. Sales-Pardo, M. Navarro, R. Guimera,

O. Yanes, Anal. Chem. 2017, 89, 3474.[69] F. Allen, A. Pon, M. Wilson, R. Greiner, D. Wishart, Nucleic

Acids Res. 2014, 42, W94.[70] H. Tsugawa, T. Kind, R. Nakabayashi, D. Yukihira, W.

Tanaka, T. Cajka, K. Saito, O. Fiehn, M. Arita, Anal. Chem.2016, 88, 7946.

[71] R. M. Salek, C. Steinbeck, M. R. Viant, R. Goodacre, W. B.Dunn, GigaScience 2013, 2, 13.

Received: April 12, 2017Published online on July 7, 2017



本文献由“学霸图书馆-文献云下载”收集自网络，仅供学习交流使用。

学霸图书馆（www.xuebalib.com）是一个“整合众多图书馆数据库资源，

提供一站式文献检索和下载服务”的24 小时在线不限IP

图书馆。

图书馆致力于便利、促进学习与科研，提供最强文献下载服务。

图书馆导航：

图书馆首页文献云下载图书馆入口外文数据库大全疑难文献辅助工具

http://www.xuebalib.com/cloud/

http://www.xuebalib.com/

http://www.xuebalib.com/cloud/


http://www.xuebalib.com/vip.html

http://www.xuebalib.com/db.php

http://www.xuebalib.com/zixun/2014-08-15/44.html


high-spatial resolution mass - download.xuebalib.comdownload.xuebalib.com/4vh4ffccrjr.pdf ·...

Documents