EXPLORING SINGLE CELL NEUROCHEMISTRY WITH MULTIMODAL MALDI MS

BY

ELIZABETH KATHLEEN NEUMANN

DISSERTATION

Submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Chemistry in the Graduate College of the University of Illinois at Urbana-Champaign, 2019

Urbana, Illinois

Doctoral Committee:

Professor Jonathan V. Sweedler, Chair Professor Martha U. Gillette Professor Mary Kraft Professor Joaquin Rodriguez-Lopez Assistant Professor Roy Dar ABSTRACT

Brain function is dependent upon the active coordination of many different cell types and subtypes at cellular resolutions. While measuring neurochemistry at cellular resolutions is important for understanding emergent properties, such as cognition and memory, analyzing the chemical nature of the brain at the single cell level is primarily limited to a handful of analytical approaches which often are only capable of measuring a specific aspect of the cell. For instance, transcriptomics measures the expression level of different genes within a cell, while mass spectrometry (MS) measures gene products and metabolites themselves. Single cells are inherently sample limited, so means of extending or expanding the amount of information that can be garnered from an individual cell is important for understanding their complex neurochemistry. Here, we combine a multitude of analytical approaches with single cell matrix- assisted laser desorption/ionization (MALDI) MS to increase the amount of information we can obtain from individual mammalian brain cells. We first developed an open source software, that simplifies microscopy-guided MALDI MS measurements and subsequent correlation of data from orthogonal approaches. Using this software, we analyzed tens of thousands of rodent cerebral cells with MALDI MS and detected over five hundred distinct lipid features. Using this metabolic information, we statistically clustered the cells in 101 unique clusters, while also finding rare lipids only present in a small fraction (<1%) of cells. Further, we extended single cell MALDI MS measurements to accommodate subsequent measurements using immunocytochemistry, infrared spectroscopy, stimulated Raman scattering microscopy, single cell transcriptomics, and capillary electrophoresis, allowing us to merge the rich metabolic information detected with MS to a variety of other systems to maximize the information that can be measured from a single cell.

ii

ACKNOWLEDGEMENTS

I would like to thank my advisor, Prof. Jonathan V. Sweedler, for the opportunity to obtain my doctorate degree within his lab and under his guidance. I appreciate his patience and support through the joys and sorrows of graduate school. I am a better person because of this experience, and I hope to use these skills and life lessons to support others and continue researching the big questions in our world. I would also like to thank the other members of my committee: Prof. Martha Gillette, Prof. Mary Kraft, Prof. Joaquin Rodriguez-Lopez, and Prof.

Roy Dar, for their guidance on and enthusiasm for my dissertation work.

I would also like to thank my collaborators who have made my dissertation work possible. Dr. Troy Comi initially trained me how to perform single cell MALDI MS and then became a coauthor on my manuscripts and a good friend. I greatly appreciate his endlessly patient hours coding, editing, and discussing research-related items. Joseph Ellis time and time again went with me for fish triangles, diet mountain dew, Boomerangs, and lunch. Without his company, graduate school would have been a lot duller, less enjoyable, and less successful.

Marina Philip is the most supportive friend I have ever had. Thank you for always listening and never minimizing my tears and fears. I could not have made it to the end without your love and support. I will never be able to thank you enough for this. Prof. Stanislav Rubakhin was always ready to listen and provided endless advice and rodent brains. Dr. Jennifer Mitchell taught me about the neuroscience behind our work and was always willing to help whenever she could.

Amelia Triplett was always happy and thankful to work on our research projects, even when I asked her to make a big table that was over a hundred pages long. I would also like to thank my other collaborators and coauthors: Prof. Bin Li, Dr. Qiyao Li, Dr. Aparna Bhaduri, Shannon

Murphy, Kisurb Choe, and Sara Bell.

iii

I would like to thank the Women Chemists Committee, Department of Chemistry

Graduate Student Advisory Committee, and First Christian Church for all the opportunities to serve my department and community. These opportunities continually reminded me of my love for science and why I wanted my doctorate degree in the first place. I especially want to thank

Dr. Lloyd Munjanja for believing in and continuing the work we have started to ensure that anyone, regardless of race, gender, orientation, or background, can work towards their own doctoral degree. I would also like to thank the Women’s Resources Center, McKinley Mental

Health, Courage Connection, Prof. Kate Clancy, and Prof. Steve Zimmerman for supporting me and giving me the resources and strength to continue my degree.

I would also like to thank my friends and family. Your endless support and love, through illness, loss, and uncertainty, have allowed me to finish obtaining my doctoral degree and become the first doctor in my family. For that undeserved gift, I am eternally grateful.

iv

“Be the person you needed when you were younger.”

-Ayesha Siddiqi

v

TABLE OF CONTENTS

CHAPTER 1: INTRODUCTION AND THESIS OVERVIEW ...... 1

CHAPTER 2: EXPLORING THE FUNDAMENTAL STRICTURES OF LIFE: NON-

TARGETED, CHEMICAL ANALYSIS OF SINGLE CELLS AND SUBCELLULAR

STRUCTURES ...... 12

CHAPTER 3: MICROMS: A PYTHON PLATFORM FOR IMAGE-GUIDED

MASS SPECTROMETRY PROFILING ...... 60

CHAPTER 4: OPTICALLY GUIDED SINGLE CELL MASS SPECTROMETRY

OF RAT DRG TO PROFILE LIPIDS, PEPTIDES, AND PROTEINS ...... 88

CHAPTER 5: SINGLE CELL MALDI MS SUPERVISED BY

IMMUNOCYTOCHEMICAL CLASSIFICATIONS ...... 125

CHAPTER 6: LIPID ANALYSIS OF THIRTY-THOUSAND INDIVIDUAL

RODENT CELLS USING HIGH-RESOLUTION MASS SPECTROMETRY ...... 156

CHAPTER 7: MULTIMODAL CHEMICAL ANALYSIS OF THE BRAIN BY

HIGH RESOLUTION MASS SPECTROMETRY AND INFRARED

SPECTROSCOPIC IMAGING ...... 222

CHAPTER 8: SINGLE CELL MALDI MASS SPECTROMETRY HYPHENATED

TO STIMULATED RAMAN SCATTERING MICROSCOPY FOR ENHANCED

CHEMICAL COVERAGE ...... 270

CHAPTER 9: INTERROGATION OF SPATIAL METABOLOME OF GINKGO

BILOBA WITH HIGH‐RESOLUTION MATRIX‐ASSISTED LASER

DESORPTION/IONIZATION AND LASER DESORPTION/IONIZATION MASS

SPECTROMETRY IMAGING ...... 292

vi

CHAPTER 1

INTRODUCTION AND THESIS OVERVIEW

1.1 Research Summary

The mammalian brain is a highly complex system composed of billions to trillions of cells actively coordinated together to create emergent functions, such as learning and memory.

Many of these cells have a unique function and can generally be subdivided into categories based on morphology, function, and chemical content. Because of the inherent chemical diversity within the brain, we use single cell matrix assisted laser desorption/ionization (MALDI) mass spectrometry (MS) approaches coupled to other techniques, such as immunocytochemistry (ICC) and vibrational spectroscopy, to study the chemical components of brain cells at single cell resolution.

1.2 Single Cell Analysis

Cells are the functional unit of biological systems, ranging from single cell organisms to humans, and are inherently responsible for many complex, emergent functions like learning[1] and memory[2] within the mammalian brain. Cell-to-cell diversity not only allows these emergent functions, but single cell changes can result in disease and disorders,[3] such as beta cell malfunction within type-1 and type-2 diabetes.[4] This inherent complexity in biological systems demonstrates a need to study them at single cell resolutions, and the ubiquitous presence of cells require methods capable of high-throughput and robust measurements.[3] Of particular interest is the mammalian brain, as it is one of the most complex and least understood organs, partly due to its chemical, cellular, and spatial heterogeneity.[5] New measurement approaches and instrumentation are required to understand these complex phenomena, particularly at the single cell level, where many of these complex processes begin.

1

From an analytical perspective, single cells are challenging due to their inherently low volumes, difficulty in sampling specific cells, diversity, and subsequent data analysis.[6] Several techniques have become the standard for exploring single cell heterogeneity: microscopy,[7, 8] transcriptomics,[910] vibrational spectroscopy,[11,12] electrophysiology,[13,14] capillary electrophoresis,[15,16] and MS[17-22]. Each of these techniques can measure a different aspect of the cell; for instance, transcriptomics analyzes the expression of genes by measuring the RNA compliment of a cell,[23] while microscopy images the morphology and often the localization of fluorescently labelled antigens.[24]

While ground breaking work has been accomplished with each of these analytical approaches, they have limitations. Microscopy, for instance, can generally determine the localization of only a few compounds within a single experiment. Transcriptomics measure the expression of many cellular components, but this information does not always correlate with the physical presence of the gene product or cellular metabolites. This raises the question: Which technique allows you to obtain the most information from a single cell? In truth, it appears that the combination of multiple techniques (or multimodal analysis) together allows us to obtain a larger amount of information from single cells than any technique alone[25] and will be a central theme of this dissertation work.

1.3 Mass Spectrometry

MS is a label-free, multiplexed analytical technique that allows for the detection of discrete compounds within many different types of systems. A mass spectrometer can be broken down into three main parts: source, mass analyzer, and detector. The source is responsible for ionizing a chemical analyte of interest allowing the chemical to be manipulated by an electric field for entry into the mass analyzer. Many different types of ionization sources are available

2 ranging from electron impact to electrospray ionization, each designed to ionize different chemical classes from different sample matrices.[26]

For most of this work, we employ matrix assisted laser desorption/ionization (MALDI)

MS because it is minimally sample destructive allowing the process to be coupled to other techniques for enhanced chemical information.[27,28] MALDI MS is a soft ionization technique capable of attomole detection limits[26] of a broad range and number of chemicals, including proteins,[29] peptides,[30,31] lipids,[32,33] and small metabolites.[34,35] The analyte of interest is dissolved within an ultraviolet (UV) absorbing, organic matrix. Ionization begins with a focused

UV laser irradiating the matrix, causing the UV-absorbing matrix to erupt outward as a cloud containing the matrix and co-crystalized analytes. The ionized matrix then transfers energy to co- crystalized analytes, causing analytes to become charged and enter the mass analyzer of the instrument.[26] Additionally, electrospray ionization (ESI) was used alongside MALDI MS for lipid identifications throughout the thesis. ESI functions by creating ions via applying high current to a solution to produce an aerosol and thus is complementary to MALDI MS.[26]

The mass analyzer is the portion of the instrument that separates the ionized chemicals based on their mass-to-charge ratios (m/z). As with sources, a plethora of analyzers can be used for determining the m/z value for a specific compound. Here, we primarily used time-of-flight

(TOF) and Fourier transform ion cyclotron resonance (FT-ICR) analyzers for fast, low-resolution and slower, high-resolution measurements, respectively. TOF analyzers function by pushing ions with the same kinetic energy through a drift tube. Because the ions have the same kinetic energy, lower m/z value ions will travel through the drift tube and hit the detector (usually a microchannel plate or secondary emission multiplier) faster than higher m/z value ions, where the time it takes to travel through the tube can be converted to an m/z value with appropriate

3 calibrations. Reflectrons are often added to account and correct for the distributions of ions that exist and pushed into the TOF drift tube. TOF instruments can be used to measure a large range of m/z values, even as high as whole viral particles.[26,36]

FT-ICR instruments start by injecting ions into the center of an ICR cell. These ions are then excited by an electric field and start to orbit within the ICR cell because of an external magnetic field that is perpendicular to the electric field. The frequency at which the ions orbit is proportional to their m/z value and is detected by the ions inducing a current on two detector plates within the ICR cell. High resolution measurements are achieved by extending the time an ion is orbiting within the ICR cell as an ion is measured multiple times. FT-ICR mass spectrometers have a higher resolving power than TOF instruments and are capable of the highest mass accuracy of all MS systems, although it is more difficult to measure ions above an m/z value of 10k and resolution falls quickly with an increase in m/z value.[36-38]

1.4 Single Cell Matrix-Assisted Laser Desorption/Ionization Mass Spectrometry

MALDI MS is an apt choice for single cell analysis as it is highly sensitive, robust, fast, and nontargeted. While MALDI MS has been used for over a decade to study chemicals within single cells,[25] recent advances have been focused on improving cell analyses throughput.

Three primary methods for high-throughput acquisition of cells have been developed: mass cytometry,[39] high-density micro-arrays for mass spectrometry (MAMS) with defined locations,[32] and microscopy-guided MS.[40] Mass cytometry employs multiatom elemental tagged antibodies that bind to proteins of interest within a cell for later inductively coupled plasma MS.[39] MAMS chips are fabricated from a scaffold covered with a hydrophobic coating, where “wells” restrict diffusion during matrix application. The wells are patterned so that the location of every well is known, and the hydrophobic coating isolates the single cells within each

4 well from contaminating other wells. MAMS were used to detect metabolites, such as adenosine triphosphate, from single yeast cells and lipids within single alga cells in the initial report.[32] MAMS has been successfully used to correlate single cell MS information with other techniques, such as

Raman spectroscopy.[41] The first paper describing microscopy-guided MALDI MS of randomly seeded cells was reported by Ong et al. within our lab. They analyzed several thousand pituitary cells to detect peptides within individual cells.[42] The second paper describing this approach was reported by Jansson et al., where they profiled the peptides within pancreatic islet cells[43] and the first iteration of microMS is described before being made publicly available in 2017 and is at the center of this dissertation work.[40]

In essence, cells are dispersed on a microscope slide that has etched fiducial markers on it. The cells are then incubated with Hoechst 33342, a fluorescent DNA intercalator, before being imaged with brightfield and fluorescence microscopy. Because the brightfield and fluorescence images are inherently overlaid, microMS can be used to automatically identify cell-like features with their respective pixel coordinates that can be translated to MS stage coordinates via labeling the stage coordinates of the fiducials. Using this approach, cells in an arbitrary pattern can be imaged quickly (~1 cell/sec and ~1 cell/ 30 sec on MALDI TOF and MALDI FT-ICR instruments, respectively) for multiplexed chemical analysis of biological systems.[40] Moreover, the use of microMS enables orthogonal techniques to be performed on the same single cell, such as MALDI MS and vibrational spectroscopy.

1.5 Overview of Thesis

Following the introduction in chapter 1, chapter 2 provides additional information for nontargeted, chemical analysis of single cells[25] and chapter 3 describes microMS in greater detail.[40] Chapter 4 details the use of microMS to study the peptide and small protein markers of

5 dorsal root ganglia cells.[44] To expand the amount of information that can be gained from an individual cell, we coupled MALDI MS with other chemically-rich and informative techniques:

ICC (chapter 5 and 6), infrared (IR) spectroscopy (chapter 7),[28] and stimulated Raman scattering microscopy (SRSM, chapter 8). Finally, we end our discussion in chapter 9 with a

MALDI imaging application to ginko leaves, demonstrating that while we spent most of our research effort on mammalian systems, our research efforts can easily be applied to other biological systems, such as plants.[45]

1.6 Future Outlook

Single cell analysis using MS is only at its beginning stages and will continually expand in the coming years. While my thesis work has aided in the exploration of single cell lipid heterogeneity and combination of this information to other chemically rich techniques, there is still much work to be done. For instance, coupling single cell transcriptomics measurements to

MALDI MS cannot be consistently performed concurrently, while still obtaining quality data from both approaches. Future work will clearly be aimed at improving the information and reproducibility of this correlated analysis. A similar statement could be applied to many of the other multimodal approaches, such as improving ICC staining post-MALDI MS and increasing the number of acquired vibrational bands in spectroscopic images. Even further, MALDI MS could be coupled to other analytical techniques not discussed here: flow cytometry, genetically- encoded fluorescent probes, electrochemistry, and many others.

While initiatives that improve multimodal analysis is an important aspect of single cell analysis, I believe that improving sampling approaches will significantly improve the information garnered by both single cell MALDI MS and subsequently correlated techniques.

Because we are enzymatically dissociating cells from brain tissue, we are removing each cell’s

6 processes and, as a result, their cell-to-cell connections. Through this process, we may also be destroying fragile cells and eliminating them from MS analysis, essentially biasing our chemical classifications towards more stable brain cells. Certainly, optimizing the dissociation procedure would allow analysis of different types of cells, but the first step is to determine if cell types are lost and, if so, what cells they are. After they are identified, measures can be taken to keep them in the analysis.

Moreover, we are determining the chemical profiles of the cell soma, as opposed to cellular process. Many cellular functions within the brain involve specific process, such as an axon, indicating that important chemical information can be garnered from them. Culturing cells, however, drastically changes the chemical composition of the individual cells and may perhaps lead to less biologically-significant chemical conclusions. This, however, has not been extensively studied and determining the chemical changes and differences among cultured cells could lead to their use (or disuse) in MALDI MS analysis, allowing us to study more than the cell soma. Additionally, incorporating single cell MALDI MS measurements into tissue systems to study cell-to-cell connections is also an important avenue.

Beyond the brain, single cell MALDI MS and multimodal analysis could also be used to study other biologically important systems and their respective disease states, such as the pancreas and diabetes. By building chemical profiles of these systems, we may begin to understand how single cells affect organ function and unravel lipid function and disfunction at cellular resolutions. I am certain that single cell MALDI MS can help us better understand many biological systems, particularly the brain, and will continue to be an active area of research.

7

1.7 References

(1) W. B. Scoville, B. Milner, Journal of Neurology, Neurosurgery & Psychiatry 1957,

20, 11-21.

(2) H. Jörntell, C. Hansel, Neuron 2006, 52, 227-238.

(3) R. Zenobi, Science 2013, 342, 1243259.

(4) G. C. Weir, S. Bonner-Weir, Diabetes 2004, 53, S16-S21.

(5) C. S. von Bartheld, J. Bahney, S. Herculano-Houzel, Journal of Comparative Neurology

2016, 524, 3865-3895.

(6) T. J. Comi, T. D. Do, S. S. Rubakhin, J. V. Sweedler, Journal of the American Chemical

Society 2017, 139, 3920-3929.

(7) F. Ryttsén, C. Farre, C. Brennan, S. G. Weber, K. Nolkrantz, K. Jardemark, D. T. Chiu,

O. Orwar, Biophysical journal 2000, 79, 1993-2001.

(8) J. W. Young, J. C. W. Locke, A. Altinok, N. Rosenfeld, T. Bacarian, P. S. Swain, E.

Mjolsness, M. B. Elowitz, Nature Protocols 2011, 7, 80.

(9) A. K. Shalek, R. Satija, X. Adiconis, R. S. Gertner, J. T. Gaublomme, R. Raychowdhury,

S. Schwartz, N. Yosef, C. Malboeuf, D. Lu, J. J. Trombetta, D. Gennert, A. Gnirke, A.

Goren, N. Hacohen, J. Z. Levin, H. Park, A. Regev, Nature 2013, 498, 236.

(10) F. Tang, C. Barbacioru, Y. Wang, E. Nordman, C. Lee, N. Xu, X. Wang, J. Bodeau, B. B.

Tuch, A. Siddiqui, K. Lao, M. A. Surani, Nature Methods 2009, 6, 377.

(11) H. Wu, J. V. Volponi, A. E. Oliver, A. N. Parikh, B. A. Simmons, S. Singh, Proceedings

of the National Academy of Sciences 2011, 108, 3809-3814.

8

(12) K. Kneipp, A. S. Haka, H. Kneipp, K. Badizadegan, N. Yoshizawa, C. Boone, K. E.

Shafer-Peltier, J. T. Motz, R. R. Dasari, M. S. Feld, Applied Spectroscopy 2002, 56, 150-

154.

(13) K. Kitamura, B. Judkewitz, M. Kano, W. Denk, M. Häusser, Nature Methods 2007, 5, 61.

(14) B. Lambolez, E. Audinat, P. Bochet, F. Crépel, J. Rossier, Neuron 1992, 9, 247-258.

(15) S. Hu, L. Zhang, R. Newitt, R. Aebersold, J. R. Kraly, M. Jones, N. J. Dovichi, Analytical

Chemistry 2003, 75, 3502-3505.

(16) T. Lapainis, S. S. Rubakhin, J. V. Sweedler, Analytical Chemistry 2009, 81, 5858-5864.

(17) X. Gong, Y. Zhao, S. Cai, S. Fu, C. Yang, S. Zhang, X. Zhang, Analytical Chemistry

2014, 86, 3809-3816.

(18) T. Fujii, S. Matsuda, M. L. Tejedor, T. Esaki, I. Sakane, H. Mizuno, N. Tsuyama, T.

Masujima, Nature Protocols 2015, 10, 1445.

(19) B. Shrestha, P. Sripadi, B. R. Reschke, H. D. Henderson, M. J. Powell, S. A. Moody, A.

Vertes, PLOS ONE 2014, 9, e115173.

(20) H. Mizuno, N. Tsuyama, T. Harada, T. Masujima, Journal of Mass Spectrometry 2008,

43, 1692-1700.

(21) F. Chen, L. Lin, J. Zhang, Z. He, K. Uchiyama, J.-M. Lin, Analytical Chemistry 2016, 88,

4354-4360.

(22) W. Rao, N. Pan, Z. Yang, Journal of The American Society for Mass Spectrometry 2015,

26, 986-993.

(23) R. Lowe, N. Shirley, M. Bleackley, S. Dolan, T. Shafee, PLOS Computational Biology

2017, 13, e1005457.

(24) V. Sandoghdar, Contemporary Physics 2015, 56, 123-143.

9

(25) E. K. Neumann, T. D. Do, T. J. Comi, J. Sweedler, Angewandte Chemie International

Edition, 0.

(26) A. Cohen, J. Banoub, Applied Spectroscopy Reviews 2009, 44, 210-230.

(27) T. J. Comi, M. A. Makurath, M. C. Philip, S. S. Rubakhin, J. V. Sweedler, Analytical

Chemistry 2017, 89, 7765-7772.

(28) E. K. Neumann, T. J. Comi, N. Spegazzini, J. W. Mitchell, S. S. Rubakhin, M. U.

Gillette, R. Bhargava, J. V. Sweedler, Analytical Chemistry 2018, 90, 11572-11580.

(29) V. Labas, A.-P. Teixeira-Gomes, L. Bouguereau, A. Gargaros, L. Spina, A. Marestaing,

S. Uzbekova, Journal of Proteomics 2018, 175, 56-74.

(30) S. S. Rubakhin, J. D. Churchill, W. T. Greenough, J. V. Sweedler, Analytical Chemistry

2006, 78, 7267-7272.

(31) C. R. Jimenez, P. A. v. Veelen, K. W. Li, W. C. Wildering, W. P. M. Geraerts, U. R.

Tjaden, J. v. d. Greef, Journal of Neurochemistry 1994, 62, 404-407.

(32) P. L. Urban, K. Jefimovs, A. Amantonico, S. R. Fagerer, T. Schmid, S. Mädler, J.

Puigmarti-Luis, N. Goedecke, R. Zenobi, Lab on a Chip 2010, 10, 3206-3209.

(33) M. A. Hossen, Y. Nagata, M. Waki, Y. Ide, S. Takei, H. Fukano, G. A. Romero-Perez, S.

Tajima, I. Yao, K. Ohnishi, M. Setou, Analytical and Bioanalytical Chemistry 2015, 407,

5273-5280.

(34) A. Amantonico, J. Y. Oh, J. Sobek, M. Heinemann, R. Zenobi, Angewandte Chemie

International Edition 2008, 47, 5382-5385.

(35) A. Amantonico, P. L. Urban, S. R. Fagerer, R. M. Balabin, R. Zenobi, Analytical

Chemistry 2010, 82, 7394-7400.

10

(36) E. R. Amstalden van Hove, D. F. Smith, R. M. A. Heeren, Journal of Chromatography A

2010, 1217, 3946-3954.

(37) R. M. A. Heeren, A. J. Kleinnijenhuis, L. A. McDonnell, T. H. Mize, Analytical and

Bioanalytical Chemistry 2004, 378, 1048-1058.

(38) Y. Cho, A. Ahmed, A. Islam, S. Kim, Mass Spectrometry Reviews 2015, 34, 248-263.

(39) D. R. Bandura, V. I. Baranov, O. I. Ornatsky, A. Antonov, R. Kinach, X. Lou, S. Pavlov,

S. Vorobiev, J. E. Dick, S. D. Tanner, Analytical Chemistry 2009, 81, 6813-6822.

(40) T. J. Comi, E. K. Neumann, T. D. Do, J. V. Sweedler, Journal of The American Society

for Mass Spectrometry 2017, 28, 1919-1928.

(41) S. R. Fagerer, T. Schmid, A. J. Ibáñez, M. Pabst, R. Steinhoff, K. Jefimovs, P. L. Urban,

R. Zenobi, Analyst 2013, 138, 6732-6736.

(42) T.-H. Ong, D. J. Kissick, E. T. Jansson, T. J. Comi, E. V. Romanova, S. S. Rubakhin, J.

V. Sweedler, Analytical Chemistry 2015, 87, 7036-7042.

(43) E. T. Jansson, T. J. Comi, S. S. Rubakhin, J. V. Sweedler, ACS Chemical Biology 2016,

11, 2588-2595.

(44) T. D. Do, J. F. Ellis, E. K. Neumann, T. J. Comi, E. G. Tillmaand, A. E. Lenhart, S. S.

Rubakhin, J. V. Sweedler, ChemPhysChem 2018, 19, 1180-1191.

(45) B. Li, E. K. Neumann, J. Ge, W. Gao, H. Yang, P. Li, J. V. Sweedler, Plant, Cell &

Environment 2018, 41, 2693-2703.

11

CHAPTER 2

EXPLORING THE FUNDAMENTAL STRUCTURES OF LIFE: NON-TARGETED, CHEMICAL

ANALYSIS OF SINGLE CELLS AND SUBCELLULAR STRUCTURES

Notes and Acknowledgements

This chapter is adapted from a manuscript coauthored by Thanh D. Do, Troy J. Comi, and Jonathan V. Sweedler, submitted for publication on October 17, 2018 and accepted

November 30, 2018 by Angewandte Chemie, DOI: 10.1002/anie.201811951. The authors gratefully acknowledge support from the National Institutes of Health, Award Number P30

DA018310 from the National Institute on Drug Abuse, and from the National Institute of Mental

Health, Award Number 1U01 MH109062. E.K. Neumann and T.J. Comi. acknowledge support from the National Science Foundation Graduate Research Fellowship Program and the

Springborn Fellowship. T.J. Comi received additional support through the Training Program at

Chemistry-Interface with Biology (T32 GM070421).

2.1 Introduction

Cell theory, first put forward in 1839 by Schleiden and Schwann,[1] originated from the concept that the cell is the basic structural and functional unit of all living organisms. Each cell is a marvel of detailed and complex architecture, with its history reflected in an inherited morphology. After observing the division of red blood cells in chicken embryos, Raspail and

Remak[2] contributed an important tenet to cell theory with the knowledge that new cells are formed from pre-existing cells. The phrase “omnis cellula e cellula” − each cell stems from another cell – was popularized by Virchow,[3] and is a more precise version of Pasteur’s terse statement on biogenesis “omne vivum ex vivo” – all life is from life. A collection of individual cells may communicate, multiply, and differentiate to form a tissue. The variety of unicellular

12 communities and the complexity of multicellular organisms require an astonishing chemical diversity to produce a broad range of functions in an organism. Chemical cues in the surrounding microenvironment govern cell division, morphogenesis, and aging, and have the ability to alter the phenotypes of even genetically-identical cells. Although the concept that disease involves changes in normal cells was proposed by Virchow in the 1850s,[4] past technologies limited most research and clinical practices to bulk analyses of macroscopically homogeneous cell populations. Cell classification based on physical traits (e.g., size, shape, color, etc.) may be insufficient to unravel the complex nature of chemical information. Bulk measurement generates a population-averaged profile that hinders the investigation and identification of low-abundance cell subtypes that are implicated in disease etiology, progression, and exacerbation. In addition, extracting analytes from a homogenized sample dilutes analytes derived from rare cells while increasing the chemical complexity of the mixture. As such, the shift from bulk to single cell analysis is inevitable.

Given the importance of single cell chemical measurements, why are they rare? Efforts to measure the chemical contents of individual cells must overcome the challenges of sampling the small quantity of chemically diverse analytes found within single cells. Due to their small volume, a technique with a femtomole limit of detection used to analyze a 1,000-µm3 cell (one picoliter; approximately the size of a typical mammalian cell) is only capable of detecting compounds at millimolar concentrations or above, which may be insufficient for most analytes.

Other limitations include the ability to isolate single cells from tissues or cell cultures, the stability of cells, and difficult cell manipulations, which often lead to experimental artifacts.[5]

The nature of cellular heterogeneity creates additional challenges for single cell measurements. In contrast to the investigation of more widespread cells, targeting uncommon

13 phenotypes requires the acquisition of larger samples to ensure a rare event with statistical confidence. Factors such as the ratio of cells to debris in the sample, the signal-to-noise ratio of detected signals to background, the frequency of rare cellular events, and the sensitivity of the instrument should all be considered. The number of total measurements depends on the desired standard error of the mean and the predicted frequency of the rare events occurring. Rare events can vary drastically from a few percent to one in several billion, as in the case of circulating tumor cells in the bloodstream of a metastatic cancer patient.[6] Researchers have devised clever approaches to overcome the challenge of finding rare cells, which include combining cell isolation/sorting schemes with analyses at single cell resolution. However, such schemes require a priori knowledge about the cells under investigation, and hence are limited to targeted analyses. Nevertheless, single cell analysis plays an essential role in modern cell biology, including measuring diversity within a population, identifying rare subpopulations, tracing cell lineages and phenotypes during normal development or disease stages, and discovering new cell types.[7]

While a few earlier examples have been reported,[8] the current era of single cell chemical analysis began twenty years ago with several advances. For example, Jorgenson and Kennedy[9] used capillary separations to profile amino acids and neurotransmitters from three different neurons of the land snail Helix aspersa. Wightman and co-workers[10] monitored the secretion of catecholamines from single bovine chromaffin cells with a carbon-fiber microelectrode, and the

Ewing group[11] estimated the free dopamine in the cytoplasm of the dopamine cell of Planorbis using voltammetry and capillary electrophoresis (CE). In 1992, Eberwine[12] performed more comprehensive single cell analysis and demonstrated that the molecular profile of a single, potentiated CA1 neuron depends on the abundance of multiple RNAs. These early examples

14 highlighted the need of single cell studies by establishing previously unknown molecular variation within individual cells.

The links between genotype and phenotype are a consequence of the intricate pathways between gene transcription, mRNA translation, and protein-level regulation. Alone, the cellular genome does not fully explain the complexity and dynamics of the cellular peptidome and metabolome, which are important for intracellular function and intercellular communication.

Therefore, while the genome and transcriptome continue to provide insights on cell function, single cell metabolomics and peptidomics (SCMP) measurement strategies are required to illuminate the identity, dynamics, and functions of key molecules and relate them to biological and physiological processes.

While SCMP approaches have contributed to numerous advancements,[13] here we limit the scope of our discussion to non-targeted techniques. As each cell is a small-volume, mass- limited sample, bioanalytical techniques have been downscaled and hyphenated to improve detection and characterization capabilities. The diverse repertoire of cells poses a tremendous challenge to obtaining distinct types of information associated with molecular processes that dictate cell-fate decisions. Matching biological questions with an appropriate SCMP technique requires balancing their competing requirements and understanding that is currently impossible to assay all chemicals present within every individual cell in a system. As illustrated in Figure

2.1, we categorize the approaches based on their ability to deeply catalog the contents of a few cells, assess the native anatomical context of cellular analytes via chemical imaging, and measure dissociated cells at high(er) throughput. In the first case, the complexity of biological samples warrants the leading role of separation techniques in a “divide and conquer” scheme to maximize the chemical information gained from each measurement. With slower analyses,

15 fractionation coupled to mass spectrometry (MS) is practically limited to smaller numbers of cells. Because of this, the cells must be chosen carefully, further limiting the use of MS as a population-discovery technique. Anatomical information is maintained with chemical imaging approaches as they can map the two-dimensional (2D) or three-dimensional (3D) spatial distribution of biomolecules. However, without a separation, they tend to detect more abundant and easily ionizable compounds. Finally, cells that are isolated from tissue first may be analyzed faster and more completely, but often at the cost of the original cellular context. Profiling measurements performed at high throughput increase the odds of observing minor phenotypic differences and detecting rare cells. When cells are isolated from each other before measurements, there are relaxed constraints on the sampling probe, allowing greater flexibility in solvent extraction protocols and matrix-assisted ionization modalities.

The outlook that native context, increased metabolic coverage, and assay speed are dependent and often mutually exclusive may not appear encouraging. However, researchers are overcoming limitations by improving instrumentation or coupling multiple analytical methods.

We largely focus on recently developed MS and MS hyphenated approaches that have pushed the limits of both the breath and the depth of cellular analysis.

To an extent, improving chemical coverage, maintaining spatial integrity, and increasing throughput are active research areas. After centuries of scientific investigation, we are linking the fundamental connections between individual cells through transport, breakdown, and formation of molecules to the emergent functions of memory, learning, and behavior. In what follows, we outline the sample preparation approaches followed by the lower throughput comprehensive approaches, chemical imaging and finally, the high(er) throughput cell profiling techniques. As modern single cell analysis offers a glimpse at exploring biological systems at unprecedented

16 resolution, we end our discussion on the outlook of non-targeted SCMP and its applications to biological research.

2.2 Sample Preparation for Single Cell Analysis

Successful single cell measurements require well-designed sample preparation protocols to ensure meaningful results. Cells from the same tissue region may require different isolation techniques. The optimal sample treatment must be selected for the analyte of interest and will determine both the methods available for analysis and which analytes will be preserved.[14] In general, single cell analyses with non-targeted SCMP techniques are performed either at the tissue level using technologies that provide subcellular spatial resolution, or on isolated cells using methods that improve molecular characterization, but at the expense of spatial information.

For subcellular imaging, the integrity and spatial organization of tissues must be maintained during the collection, storage, and treatment procedures prior to sectioning. A recent review by

Chughtai and Heeren[15] listed common biological sample handling protocols for MS analysis, with an emphasis on sample storage, that minimize tissue deformation and degradation. For dissociated single cell samples, protocols for whole tissue digestion should be optimized to maintain the viability of fragile cells while removing connective matrices that may hinder the study of individual cells. The suitability of treatment options is often evaluated on a per-case basis, with attention to maintaining endogenous distributions of analytes.

While direct tissue analysis places a stringent requirement for imaging probes to provide subcellular spatial resolution, recent advances have focused on sampling dissociated single cells at the expense of native spatial information. Manual cell isolation remains effective when examining a limited number of cells. Single cells with known features, morphologies, or precise anatomical positions can be isolated and then placed manually or seeded randomly on the

17 analysis surface. For example, the well-annotated central nervous system of Aplysia californica enables manual isolation with sharp tungsten needles after enzymatically degrading the connective tissues. After years of morphological, electrophysiological and biochemical research, neurons in many clusters have been characterized.[16] Unfortunately, many mammalian cell types are challenging to identify without the aid of staining techniques, such as immunohistochemistry, which are incompatible with most MS techniques.

On the other hand, enzymatic treatments are applicable to tissue types where some cell loss during isolation is not a major concern. For example, neurons can be dissociated and cultured in a few hours from dorsal root ganglia (DRG), and remain alive and growing for weeks. In general, enzymatic dissociation increases throughput and mitigates the expertise required to perform manual isolation for cell systems that can withstand more rigorous treatments. Each dissociation protocol involves the combination of proteases (e.g., collagenase, papain, trypsin, etc.), inert proteins (e.g., bovine serum albumin), balanced salt solutions, and other added ingredients to preserve or stabilize cell contents. Once dissociated, cells can be cultured, sorted with microfluidic devices, or simply deposited on a substrate for MS analysis. A recent review by Hosic et al.[17] is dedicated to microfluidic sample preparation for single cell analysis. In many cases, the use of cell-sorting devices is optional, as recent methodologies make it possible for thousands of dispersed cells and non-cellular targets to be assayed in one experiment, as discussed in the following sections.

2.3 Comprehensive Multimodal Chemical Analysis

Separation techniques have historically played a key role in both preparative and analytical cellular level studies as they ensure the necessary selectivity and chemical purity as well as reduce isobaric interferences. The retention time within a chromatographic column

18 provides a metric for compound identification. Following fractionation, several detection modalities other than MS are available, including electrochemistry, ultraviolet–visible and laser- induced fluorescence (LIF) spectroscopies, among others. LIF generally requires derivatization to attach a fluorophore, unless the analyte of interest is intrinsically fluorescent. While derivatization is sometimes impractical, fluorescence detection has the lowest limits of detection, at times less than a hundred molecules.[18] Electrochemical detection provides attomole detection limits[19] without derivatization and can be miniaturized without sensitivity losses.[20]

Nonetheless, de novo identification of chemical analytes via optical or electrochemical methods is difficult due to the lack of chemical resolution.[20]

While fundamental chemical information from single cells has been obtained by single cell liquid chromatography (LC),[21] most recent single cell separation experiments are performed with miniaturized CE systems that require only nL to pL quantities of solution and are suitable for analyses of whole cells, fractions of cells, or cellular extracts.[22] CE separates compounds based on differences in their electrophoretic mobility, which depends upon the number of charges, size, and shape of each compound.[23] A whole cell or portion of a cell can be manually injected onto the column using a microscope and micromanipulator with subsequent injection into the capillary using pressure or electroosmotic flow and lysis. Alternatively, analytes are extracted from the cell and injected directly into the capillary.[24]

CE analysis of cultured cells is especially challenging as cells can adhere to culture substrates, making them difficult to remove.[22] Mechanical or enzymatic removal of cells may lead to deformation and release of important compounds. The Allbritton laboratory[25] circumvented the challenges of mechanical isolation with a pulsed laser lysis system, allowing immediate capillary loading that reduces changes to cell physiology. An alternative, rapid lysis

19 scheme utilizes an electrode to induce a 30 V potential difference across the cell membrane.[26]

To improve the throughput of single cell CE, many laboratories have coupled microfluidic devices to the front end of a CE system.[27] For example, a microfluidic device was used for cell lysis and CE separation before flowing through an ESI emitter into the mass spectrometer, increasing the analysis rate to 12 cells per minute.[27b]

MS remains an effective choice for non-targeted analysis because of its mass selectivity, robustness, sensitivity and capability to identify unknown structures. MS is further enhanced when combined with separations for multi-modal analysis. Solvent-assisted ionization methods, such as electrospray ionization (ESI), help preserve the separation efficiency of capillary electrophoresis (CE) while achieving sensitive MS analysis.[28] Non-targeted CE-MS analysis has been successfully used to probe the metabolic contents in a variety of biological systems, such as neurons,[29] cancer cells,[30] plants,[31] and red blood cells.[32] CE-LIF and CE-ESI-MS of individual cells remain active areas of research, with samples ranging from developing embryos within the species Xenopus laevis[33] to human stomach cancer cells.[34] Recently, the Nemes group[35] quantified proteins using bottom-up CE-ESI-MS and detected over 438 non-redundant protein groups during different developmental stages of the X. laevis embryo. Further, they studied the metabolic contents within certain blastomeres that contribute to cell differentiation into neuronal, epidermal, and hindgut tissues within 16-cell X. laevis embryos. Overall, 80 metabolites were detected and used to develop predictive metabolic profiles for undifferentiated cells. This same group further investigated temporal changes by utilizing live-cell CE-MS to probe metabolic differences between the dorsal and ventral side of X. laevis embryos at eight- and sixteen-cell stages. Metabolic differentiation occurred as early as eight cells, which is earlier than originally predicted.[35]

20

Other metabolic investigations include quantitation of mono-, di-, and tri-nucleotides, along with other energy-related anionic molecules within single Aplysia R2 neurons. The energy balance within single cells from such measurements was found to be comparable to that obtained by bulk sampling.[36] CE is applicable to the separation and quantitation of isobaric metabolites, namely L- and D-amino acids. The Sweedler group incorporated sample stacking with single cell

CE-LIF to allow chiral separations of the D- and L-forms of aspartate and glutamate within individual Aplysia neurons.[37] The separation power and large dynamic range of CE allowed statistically distinct levels of the D-amino acids to be quantitated within different cell types. CE-

LIF has also been used to study proteins within individual cancer cells, discovering heterogeneity within the expression of cysteine cathepsins.[38]

While single cells are the smallest functional unit of life, subcellular organelles actively manage and participate in hierarchical organization, expression and intercellular communication.

Within a decade of the first single cell analysis, CE progressed to enable investigation of structures as small as single vesicles. As early as 1998, Lundqvist et al.[39] extracted single, secretory vesicles from the atrial gland of A. californica and utilized CE-LIF to detect abundant amounts of taurine, supporting the role of taurine as a neuromodulator or hormone. The

Allbritton group[40] later applied their laser system to selectively disrupt a portion of a cell, allowing selected CE-LIF analysis of specific portions of cell membranes and processes for localized detection of chemicals. Another strategy for subcellular analysis utilized on-column treatment of cell membranes with digitonin to selectively remove nuclei for CE-LIF analysis.[41]

The combination of trypsin and digitonin treatment allows on-column isolation of mitochondria for CE quantification.[42]

21

Single cell CE analyses have successfully targeted diverse biochemical compounds from an eclectic collection of organisms. CE will continue to benefit from improvements in sampling throughput, as well as the sensitivity of the various detection modalities. An exciting prospect for single cell CE is the repeated analysis of living single cells to monitor biological phenomena such as cell development or reactions to stimuli. In 2017, Nemes and colleagues[33a] used CE-

ESI-MS to study the metabolic changes in the developing X. laevis embryo. They probed the

V1R cell twice and obtained results comparable to a CE-MS analysis of the dissected cell

(Figure 2.2A). Of note, they monitored metabolic changes during formation of a neural cell lineage (Figure 2.2B) and by continually analyzing the progenitor cell, detected 100 molecular features that displayed significant changes. For instance, daughter cells contained decreasing amounts of aspartate and increasing amounts of GABA. Coupling multiple techniques improve overall chemical information and is another exciting avenue for probing live cells. The Sweedler group[43] performed CE-MS analysis on thalamic neurons after patch clamp electrophysiology was used to characterize the electrical activity of selected cells. Figure 2.2C displays the cell recordings and metabolic profile from a ventral basal thalamocortical neuron as compared to the thalamic reticular nucleus shown in Figure 2.2D. Over 100 different metabolites were detected within the samples. Analysis of living single cells allows for dynamic chemical analysis, which will certainly provide a better understanding of many metabolic processes, particularly in disease states or developmental stages.

Given the importance of mapping individual cell phenotypes, separation of cell contents maximizes the depth of cellular profiling, albeit, at the expense of cell throughput. Continued improvements in scalability will enable proportional increases in the breadth of these experiments. CE-MS is especially advantageous as the orthogonal mass separation ameliorates

22 degraded resolution during fractionation. Beyond accelerating separations, preparation and manipulation of cells or organelles prior to fractionation help stratify phenotypically distinct cells for comprehensive analysis, greatly improving the quality of data obtained in a limited time frame. From this aspect, antigen and microfluidic-based sorting are natural complements to the in-depth analysis provided by CE.

Another strength of CE is its flexibility in performing several types of separations, which can target specific biochemical classes or separate enantiomers and other kinds of similar structures. Clever additives and covalent modifications will allow separation of currently unresolved peaks. Specific to single cell analysis, automated sorting and repeated analyses will likely find more applications and be adapted to commercial systems. Continued work with multimodal analyses will help streamline fractionation and optimize workflows, leading to more complete descriptions of single cells from smaller samples than ever before.

2.4 Imaging Methods for Chemical Analysis of Single Cells

Individual cells often have distinct roles that depend on their position within a tissue.

Chemical imaging allows for recognition and identification of cells within their biological context. A prominent, non-targeted chemical imaging modality is MS imaging (MSI), which we will focus on here. MSI methods raster a desorption probe across a tissue sample at cellular spatial resolutions to acquire mass spectra from cells while recording their positions within the context of their native environment. The resulting images ideally represent the spatio-chemical distribution of ionizable compounds in the tissue. Realistically, the limit of detection determines observable analytes and is affected by instrumentation, sample complexity and other factors.

Secondary Ionization Mass Spectrometry (SIMS) Imaging

23

One of the oldest chemical imaging methods used to obtain cellular resolution is SIMS, which uses a primary ion beam that can achieve submicron footprints to sputter a sample surface.

The collisions of primary ions with the sample surface cause ejection and ionization of analytes that are subsequently detected by a mass analyzer. Traditionally, SIMS analyses are categorized into two operating modes: dynamic and static.[44] The commercial CAMECA SIMS instruments were developed for dynamic SIMS analysis, with primary ion sources focused down to 50 – 200 nm probe diameters while maintaining high energy fluence. The term “NanoSIMS” was coined for many dynamic SIMS instruments, including those commercialized by CAMECA. Static

SIMS is usually described as being less destructive due to its lower primary ion dose (below 1012 ions/cm2). Originally used for analyzing organic molecules on the surface monolayer, static

SIMS is becoming widely used in biological MSI.[45]

SIMS imaging has been used to probe halogens, enriched metals, and isotopically labeled additives;[46] measure the cellular uptake and distribution of drugs in different co-cultured cell types;[47] and visualize different stages of cell cycles.[48] NanoSIMS has found wide application in biological geochemistry, cell biology, and microbiology to study the uptake, assimilation, storage, and translocation of trace elements.[49] A recent review by Musat et al.[49] highlighted the use of NanoSIMS-based methodologies to identify, quantify and visualize the incorporation of labeled substrates, many of which were performed at single cell resolution. In one pioneering

[50] work, Lechene et al. used NanoSIMS imaging to quantify N2 fixation by individual bacterium inhabiting the gills of the ship-worm Lyrodus pedicellatus (i.e., single bacterium within an animal cell). Popa et al.[51] used NanoSIMS to characterize cellular development and metabolite exchange in and between individual vegetative and heterocyst cells of Anabaena oscillarioides, a filamentous freshwater cyanobacterium. The two cell types of A. oscillarioides coexist in the

24 same filament, but only heterocyst cells specialize in nitrogen fixation. Other vegetative cells

13 15 participate in oxygenic photosynthesis and CO2 fixation. By adding tracer-levels of C and N, tracking stable isotopes from inorganic pools to their cellular fate was achieved with 100-nm spatial resolution using NanoSIMS. In a similar study, Kuypers and colleagues[52] measured the

13 - 15 + assimilation of H CO3 and NH4 by individual cells of the anaerobic, phototropic bacterial species Chromatium okenii, Lamprocystis purpurea, and Chlorobium clathratiforme. The study revealed a large range of uptake rates of ammonium and inorganic carbon for single cells of the same species, which might result from genomic diversity in phylogenetically identical but physiologically distinct populations.

SIMS can be combined with other single cell approaches. In environmental microbiology, NanoSIMS has been combined with fluorescence in situ hybridization (FISH),[53]

Raman spectroscopy,[54] and high resolution optical microscopy.[55] McGlynn et al.[53a] coupled

FISH and NanoSIMS to investigate the metabolic activities of single archaeal and bacterial cells within a consortium (see Figure 2.3A). Saka et al.[55a] correlated stimulated emission depletion microscopy with NanoSIMS imaging to visualize and quantify the turnover of isotopically labeled proteins in different organelles from cultured hippocampal neurons. These are a few of many examples where the combined methods allow simultaneous activity measurements and identification of single microbial cells.

NanoSIMS has recently been applied in the field of biomedical imaging. Nolan and co- workers[56] employed mass cytometry labels (e.g., affinity probes with metal isotope tags) in an imaging application referred to as multiplex ion beam imaging (MIBI). MIBI has several advantages over conventional immunohistochemistry techniques, including the absence of background auto-fluorescence signals, an order of magnitude larger dynamic range, and a higher

25 plexity of up to 60 targets. In a recent report, Angelo et al.[57] used MIBI to image human breast tumors and revealed immunophenotypes of cell subpopulations that could be related back to the original clinical pathology of the tissue (Figure 2.3B). Another recent, exciting development is the use of atomic recombination to measure distances that are smaller than instrumental imaging resolution. Atomic recombination occurs when atoms from different molecules combine to form diatomic ions and is sensitive to the pairwise distance between molecules. Moss and Boxer[58] exploited the atomic recombination of 13C and 15N to form 13C15N- to assess the lateral distribution heterogeneity of lipids at a spatial resolution of 100 nm. The recombination phenomenon may eventually allow probing the proximity of a lipid to a protein of interest at nanometer-length scales.

While biological information has been and will continue to be gained from NanoSIMS, time-of-flight (TOF)-SIMS instruments are suitable for non-targeted analysis because of improved detection of intact molecular ions over a larger mass range. Rather than mapping internal, atomic species using dynamic mode, early TOF-SIMS experiments were performed in static mode and focused on the identification and localization of cell surface compounds.

Pioneering work by the Winograd and Ewing groups[59] developed the unique strength of SIMS in biological imaging of single mammalian cells and intact tissues. The authors used a 15-keV

In+ liquid metal ion beam to examine the lipid distribution along the conjugation junction of mating Tetrahymena, a model for studying membrane fusion. The images revealed a decrease in abundance of phosphatidylcholine and an increase in 2-aminoethylphosphonolipid at highly curved fusion pores, suggesting that Tetrahymena direct lipids to adjust membrane structure during conjugation. Shortly after, Monroe et al.[60] studied the subcellular localization of vitamin

E in the soma-neurite junction of single, isolated neurons from A. californica using a TRIFT III

26 instrument equipped with a 22-keV Au+ liquid metal ion source. Using the same instrument,

Tucker et al.[61] imaged the cultured neurons under different treatment conditions and mapped the distributions of cholesterol, vitamin E, and phosphocholine head groups in cell soma, neurites, and growth cones.

+ + The development of polyatomic ion beams, such as the C60 and Ar1000-4000 , marked a

+ new era for TOF-SIMS imaging of biological samples. The C60 ion beam developed by the

Vickerman group and Ionoptika[62] has higher high mass ion yields, molecular depth profiling

+ + and reduced subsurface chemical damage compared to early Ga and SF5 ion sources. Advances in instrumentation have improved the compromise between high mass resolution and high spatial resolution, while further reducing analysis time.[63] They modified the Ionoptika J105 3D

+ Chemical Imager by equipping the instrument with a 40-keV C60 ion gun with a minimum spot size of 200 nm. The capabilities of the instrument were demonstrated by performing 3D imaging of single benign prostatic hyperplasia (BPH) cells and 2D imaging of single HeLa, human cheek cells, and sectioned Xenopus blastomers (0.8 – 1.3 mm in diameter). The images of single cells revealed the distributions of adenine and lipids. The tandem MS capability of the J105 allowed the identification of several intact lipids.[63] Passarelli et al.[64] and Lanni et al.[65] further demonstrated the capability of TOF-SIMS with tandem MS in mapping various intact lipids across the cell surface of single neurons of A. californica.

The use of cluster ion sources has demonstrated great potential for molecular 3D imaging of single cells (for a recent review on 3D imaging with SIMS, see[66]). A dual ion source system was shown to be useful in reports by Breitenstein et al.[67] and Nygren et al.,[68] who performed

3D TOF-SIMS imaging of normal rat kidney and anaplastic thyroid carcinoma cells,

+ respectively, at a spatial resolution of 300–350 nm using a 25-keV Bi3 liquid metal ion beam for

27

+ + imaging and a 10-keV C60 for etching and removing the damaged layers caused by the Bi3 beam. A variety of endogenous amino acids, cholesterol, and intact phospholipids were detected

(see Figure 2.4). Castner and colleagues[69] used the same ion beam combination but at higher

3+ 2+ energies (25 keV-Bi and 20 keV C60 ) to image NIH/3T3 fibroblasts in 2D and 3D. Brison et al.[70] applied a similar dual beam mode with the sputter depth calibrated using atomic force microscopy to perform 3D imaging of native and non-native chemical species in HeLa cells.

Fletcher et al.[71] performed 3D TOF-SIMS analysis of individual oocyte cells mounted

+ on copper tape using a single 40-kV C60 beam. Aside from common ions such as cholesterol and oleic acid, they observed groups of mass spectral peaks occurring at m/z 540–700 and m/z

800–1000 that might originate from glycosphingolipids of the membrane constituents, although unambiguous identifications were not possible. Nonetheless, the signals at low mass, such as phosphocholine and adenine, produced 3D visualizations of the membrane and nucleus, as demonstrated by Vickerman and co-workers[72] for HeLa-M cells. The localization of small molecules and metabolites in single bacterial cells was recently reported by Tian et al.,[73] who showed direct localization of unlabeled tetracycline and ampicillin in single, antibiotic-dosed E. coli cells (about 2 µm long and 0.25–1.0 µm in diameter) using TOF-SIMS 3D imaging at 300 nm spatial resolution.

MALDI MSI

While SIMS imaging offers superior spatial resolution, it achieves a limited mass range because the ion beams can cause fragmentation. A complementary imaging approach, MALDI

MSI, is a soft ionization method that can ionize large proteins, although the use of a matrix can cause other issues. MALDI MS directly probes analytes from a solid sample, such as dispersed cells or tissue sections,[74] and is a highly sensitive analytical technique capable of attomole

28 detection limits.[75] MALDI MS is uniquely suited for single cell analysis because it can probe cells directly with minimal perturbations and can analyze low amounts of many chemical compounds. Most of the detected analytes are singly-charged positive or negative molecular ions.[76] The sample is prepared by “dissolving” analyte compounds in the matrix, typically a small, organic compound that absorbs UV light. Irradiating the sample with a focused UV laser causes the matrix to erupt and generate gas-phase matrix and analyte ions representative of the ablated area. In addition, MALDI is more tolerant of high concentrations of salt as often found in biological media.[77] The robustness of MALDI MS is particularly appreciated in the analysis of complex biological specimens, where sample treatments frequently fail to remove all analyte interference.

Matrix application is critical for MALDI MS as the nature of matrix compounds and the methods used to apply the matrix determine the chemical coverage, analyte migration, and the overall quality of measurements. Several strategies for applying the matrix compound to the sample have been established, including sublimation,[78] nebulization,[79] inkjet printing,[80] and electrospray.[81] Each MALDI MS measurement strikes a balance between extraction efficiency and analyte diffusion, which is especially problematic for tissue imaging.[79] The choices for

MALDI matrices and their application continues to expand as novel compounds are found to be suitable matrices for specific applications or classes of analytes. MALDI MS provides broad chemical coverage of analytes, including proteins,[82] peptides,[83] lipids,[84] and small metabolites.[85] A frequently cited limitation is isobaric interference from the MALDI matrix, which complicates metabolite detection without high-resolution mass analyzers. While MALDI

MS is a destructive technique, it often consumes only a fraction of the sample,[86] allowing repeated or follow-up analyses to enhance the amount of information acquired from the same

29 target.[87] This is especially useful for prescreening cells to determine rare cells or to classify cells into subpopulations based on representative markers for subsequent analysis. Cellular and subcellular imaging (<10 µm) with MALDI MS is still not common using most commercial instruments. Laboratories have achieved single cell and subcellular analysis of tissue samples by using custom instrumentation capable of significantly smaller spatial resolution (several microns). High spatial resolution is dependent not only on the laser spot size, but also matrix crystal size and the sensitivity of the mass spectrometer. In 2010, 5-µm spatial resolution mouse bladder tissue imaging was accomplished by the Spengler group.[88] In 2012, Schober et al.[74b] developed an atmospheric MALDI source with 7-µm spatial resolution, capable of subcellular lipid and metabolite localization within HeLa cell cultures. Only months later, the Caprioli laboratory[89] demonstrated a 1-µm beam spot size achieved with transmission geometry illumination. The source arrangement was used to image individual HEK-293 cells and detected insulin differences between nuclei and cytoplasm within human pancreatic islets. More recently, in 2017 the Spengler group[90] developed an atmospheric pressure MALDI imaging source capable of 1.4-µm lateral resolution. The source capabilities were demonstrated by probing the subcellular lipid, metabolite, and peptide chemical differences between different sections of the

Paramecium caudatum cellular membrane with a 3-µm pixel size (Figure 2.5A–F).

Alternative Imaging Approaches

While SIMS and MALDI MSI are the most common MSI techniques capable of cellular imaging, alternatives are being developed. The Zare group[91] developed laser desorption/ionization droplet delivery mass spectrometry, which is capable of 3-µm spatial resolution, to image tissue samples as well as measure both single cell apoptosis and live cell exocytosis. Laskin’s group[92] has been pushing the limit of NanoDESI, recently achieving a

30 spatial resolution of 11 µm. In another example, Yang and colleagues[93] developed single-probe

MS, which is effectively a miniaturized probe constructed from a fused silica capillary with an

ESI emitter placed inside a dual-bore needle. The dual-bore needle can puncture cell membranes, gaining direct access to cell cytoplasm. Single-probe MS was first used to interrogate individual

HeLa cells and detected many small molecules and lipids within individual cells. Single-probe

MS has since been used within an imaging context, achieving an 8.5-µm spatial resolution,[94] allowing image acquisition of lipids within single cells. Single-probe MS offers the highest spatial resolution among solvent-based, ambient approaches.

Although there are many options when choosing a chemical imaging modality, the selection of an ionization technique requires balancing the constraints of the required spatial resolution, mass range of interest, instrument availability, and cost. While SIMS has superior spatial resolution compared to other imaging modalities, most systems are too cost prohibitive to be owned by a single lab, and fragmentation prevents analysis of peptides, proteins, and larger nucleic acids. However, the “out-of-box” capabilities of SIMS imaging are impressive, and are becoming even more so. MALDI MS systems are more readily available, even if many of the above examples require modifications, and can more easily analyze higher mass compounds without fragmentation. Achieving the spatial resolution required to analyze single cells in the context of tissues is currently performed by a handful of laboratories using either MALDI MS or ambient ionization.

2.5 Single Cell Profiling

Performing subcellular resolution MSI introduces a host of challenges for sample preparation and instrument development, and is generally slow. An alternative method for probing the contents of single cells is to isolate them from their microenvironment and

31 individually interrogate each cell. While the spatial information is lost, profiling single cells is often faster and has better limit of detection than conventional MSI at single cell resolution.

Commercial instruments can be used without alteration, providing access to single cell measurements to more researchers. Moreover, the latest developments allow the interrogation of thousands of cells for population studies and analysis of cell heterogeneity.

Electrospray Ionization

Probe electrospray ionization (PESI) directly inserts or touches the sample with an ESI emitter, effectively performing a surface extraction. Following extraction, the emitter (often a stainless steel needle) is repositioned directly in front of a mass spectrometer inlet for ESI analysis without further addition of liquid.[99] PESI allows live cell sampling as well as enrichment of metabolites by probing the sample multiple times before MS analysis. The Zhang laboratory[100] used the approach to analyze the outer and inner epidermal cells from Allium cepa.

Several chemical differences were noted between the cell types, including a higher diversity of fructans localized within the inner epidermal cells and distinct lipid profiles.

In 2018, the Vertes group[101] combined fluorescence microscopy with capillary microsampling to perform ESI analysis of cells at different mitotic stages. The Huang lab[102] used a patch clamp capillary as an ESI emitter to study metabolic changes within single mammalian neurons after electrophysiological analysis. They determined that cells that had unusual patching profiles also had unusual chemical profiles, demonstrating the need to measure both physiology and chemistry on the same, living neuron, perhaps multiple times.

A related method, “live-cell MS” developed by the Masujima group,[103] utilizes a sharp nano-ESI emitter to puncture a cell, extracting contents (even from an organelle) into the emitter.

The emitter is then transferred to the inlet for MS analysis. Due to the size of the probe,

32 extraction does not appreciably perturb a living cell, allowing repeated sampling of the same cell.

In a typical experiment, cells are observed under a microscope and when an event of interest occurs, such as a physical response to stimuli, the nano-ESI emitter is placed into the cell for extraction. While this technique allows for high selectivity of the analyzed region of the cell, it requires fine manipulations. The Masujima group has detected over 700 analytes within an individual cell with live-cell MS[103] and applied live-cell MS to study the metabolites within plants.[104] Similarly, the Huang lab[105] demonstrated the analysis of proteins from live cells without destroying the cell. The Laskin group[106] used localized electro-osmotic extraction to drive picoliter volumes out of live cells into a nanopipette compatible with nano-ESI analysis.

They detected more than 50 metabolites, including sugars and flavonoids, some of which could be quantified using sequential extraction of a known volume of an aqueous solution containing deuterated standards. These approaches hold great promise in measuring the changes in chemical content of the same cell as a function of disease or growth. Nevertheless, it is critical to carry out parallel measurements to monitor cell conditions, and to avoid mechanically-induced changes in cell states.

MALDI MS

MALDI MS is the primary analytical technique capable of obtaining direct chemical information from cells and has been used more than the other approaches. Here we divide

MALDI MS approaches into lower throughput manual isolation and higher throughput acquisitions. Until the last decade, single cell MALDI MS required a combination of manual cell isolation, placement, and MS acquisition. Two methods for high-throughput acquisition of single cells were recently reported: high-density microarrays for mass spectrometry (MAMS) with defined locations (Figure 2.6A)[84a] and image-guided analysis of cells randomly seeded on

33 a transparent substrate (Figure 2.6B).[107] Each method offers unique performance advantages and enables the analysis of hundreds to thousands of cells within a single experiment.

Direct MALDI MS Profiling

The Greef group[95] reported the earliest examples of single cell MALDI MS in 1993, where they profiled peptides from individual Lymnaea stagnalis neurons. Improvements to mass spectrometer sensitivity allowed detection of more analytes from single cells and even tandem

MS for structural identification without a priori knowledge. Important neuroscience findings have direct roots in the single cell MALDI MS experiments pioneered by Greef, Geraerts,

Sweedler, Burke, and others.[16b, 83b, 96] Despite the limitations of this early research, hundreds of biologically important neuropeptides, neuromodulators, and neurotransmitters (among many other molecules) were discovered and correlated with physiological function. Discoveries were performed using samples ranging from large invertebrate neurons to small peptidergic mammalian cells.[83a, 97] One of the most recent studies involving low throughput MALDI MS was performed in 2018 by the Setou lab,[98] where they measured phosphocholine lipids in single neurons as they extended neurites.

Microarrays for Mass Spectrometry

The first single cell profiling using MAMS was reported in 2010 by the Zenobi group[84a] where they detected metabolites, such as adenosine triphosphate and uridine diphosphate glucose, from single Saccharomyces cerevisiae cells and lipids within single Chlamydomonas reinhardtii cells. MAMS chips are fabricated from a conductive slide covered with a hydrophobic coating. Wells, approximately 100 µm in diameter, are patterned by laser ablation of the coating. Each well contains a small volume that restricts diffusion during matrix application. When cells are added to the surface, they settle only within the wells. Since the wells

34 are located at known coordinates, sampling proceeds at rates of about two samples per second.

Additionally, restricting each cell to a known address simplifies correlation between different analyses, such as native fluorescence and Raman spectroscopy (Figure 2.6C). In a follow-up study, the Zenobi group[75] characterized metabolites within yeast cells, facilitating investigation of metabolic heterogeneity at the single cell level. MAMS has since been applied to a variety of cell systems, such as human blood peripheral mononuclear cells[108] and Haematococcus pluvialis.[109]

Microscopy Guided Profiling

Ong et al.[107] reported high-throughput, optical microscopy-guided MALDI-TOF MS of randomly seeded cells in 2015. They analyzed several thousand pituitary cells, frequently detecting pro-opiomelanocortin peptides. The protocol begins by incubating cells with Hoechst

33342, a fluorescent DNA intercalator, and acquiring a fluorescence image to locate the cells over the area of a microscope slide. Using the low background inherent in fluorescence images, the location of each cell is determined automatically. Registering fiducial marks that are visible in both the microscopy image and instrument camera enables the conversion of pixel positions into stage coordinates to automate the MALDI-TOF MS analysis. Knowing the location of each cell allows for mass spectral acquisition at defined points at high throughput.[107] After the initial demonstration of this approach, Jansson and coauthors[110] performed additional studies of single cells isolated from pancreatic islets of Langerhans and detected peptide hormones unique to cell types within the islet.

To expand optically guided single cell profiling beyond MALDI-TOF instrumentation,

Comi et al.[111] developed the open source software, microMS, for analyzing microscopy images and correlating cell locations to instrument positions. The software provides automatic cell

35 finding and filtering by attributes such as size, fluorescence intensity, and distance between cells, and is simple to tailor to a variety of instruments, expanding the availability of single cell profiling. The ease of switching between output coordinate systems also facilitates analysis of the same cell on multiple instruments; an example being MALDI-TOF MS analysis with follow- up CE-MS analysis (Figure 2.6D).[111] Do et al.[112] recently used microMS to set up sequential

MALDI MS analyses of rat DRG cells. Lipids, peptides, and small proteins were detected in the same cells, many of which were not previously detected in tissue homogenates or releasates. The

DRG populations were stratified with multivariate statistical analysis using peptides and proteins as potential markers. A similar method utilizing flexImaging software from Bruker Corp. was recently published by the Caprioli lab[113] to determine heterogeneity within cultured macrophage cells using a Bruker solariX XR FT-ICR mass spectrometer. The authors imaged single cells by rastering across the slide and comparing the ion images with optical images acquired beforehand.

Ultimately, they were able to show changes in lipid expression of macrophages upon chemical stimulation.

The optically guided single cell profiling approach was recently demonstrated for SIMS using microMS software.[111] In the study, Do et al.[114] utilized ionic liquid matrix-assisted 20-

+ keV C60 TOF-SIMS to profile metabolites and intact lipids in three cell types: large A. californica pedal neurons, medium-sized cell bodies of rat DRG, and small rat cerebellar cells at the rate of 600 cells/h. The single cell mass spectral datasets were subjected to multivariate statistical analyses, such as principal component analysis and t-distributed stochastic neighboring embedding, which helped distinguish cell types with similar lipid profiles, including DRG and cerebellum. In addition, these analyses suggested that lipid ratios could be endogenous markers

36 to define brain regions. Overall, advanced statistical analysis strategies are necessary to fully harness the utility of high-throughput data.

The performance metrics of SIMS and MALDI MS imaging carry over to single cell analysis, as the limitations and advantages are inherent to the ionization technique. Moreover, the

MALDI MS and SIMS examples discussed here largely sacrifice cellular connections, which can be consequential when the spatial organization of the system determines the function of a cell, such as the brain. To circumvent this loss of information, cells and organelles can be manually isolated from selected cells. While manual dissection allows for cellular even subcellular chemical analysis, it is laborious and low throughput. ESI-based methods do not always require removal from tissues, although these methods are often laborious and low throughput.

Furthermore, ESI also benefits from multiply charged ions, improving detection of proteins of higher mass, but is less salt tolerant than MALDI or SIMS. As the field of single cell profiling progresses, each approach, whether it be low throughput with known spatial location or high throughput without spatial information, will need to be developed further for adoption in general labs for fundamental analysis of poorly understood systems, such as the brain, plant tissues, or disease states.

2.6 Summary and Outlook

While many approaches can provide information on individual cells, we have focused on methods that are capable of nontargeted chemical analysis. Single cell chemical analysis has evolved rapidly over recent decades to provide ‘omics-scale molecular information at the individual cell level. Front-end techniques such as CE and LC have been downscaled to enhance separation on minute samples, and in some cases, to perform repeated sampling of the same cell for real-time metabolomic information.[24, 33a, 115] MAMS and optically-guided MS profiling now

37 facilitate the assay of several thousands of cells within a couple of hours,[75, 110, 114] while cellular and subcellular spatial resolution chemical imaging can be achieved by selected MS probes such as NanoSIMS,[49-53, 57] SIMS[44c, 63, 66-67, 71-73, 116] and MALDI.[90] Future work will be aimed at achieving faster analyses with higher mass and spatial resolution and developing more sensitive instrumentation. Several groups have achieved detection of 10 zeptomoles of material with modern MS instruments, equivalent to about 6000 molecules. Fluorescence detection can detect single selected molecules. These reported detection limits for MS should be sufficient for detecting most compounds within a single cell or most subcellular organelles. Why then is the scientific community unable to realize the limits of detection in ideal systems when applied to single cell chemical analyses? We attribute the lowered performance largely to limitations caused by the inherent chemical complexity of cells and less than ideal sampling protocols.

Obtaining isolated single cells from a tissue requires dissection, dissociation, plating, rinsing, and finally, analysis. During every single step, different compounds are potentially released or removed from cells, reducing the number of molecules remaining for detection, as well as introducing other non-native contaminants. Sampling issues are compounded by ion suppression effects, transmission efficiency through the spectrometer or column, and the native background.

As a final consideration, complex systems, such as single cells, require a high dynamic range to detect molecules at abundances over many orders of magnitude. The exciting aspect is that newer instruments have the necessary performance requirements. In addition, at least for some analytes, experiments are being performed that sample living cells within their native environment, ameliorating several of the issues related to sampling. Techniques such as CE and live-probe MS allow for live cell analyses and could be applied to native chemical analysis.

Repeated sampling of individual cells can allow monitoring of the progression of disease, such

38 as cancer and diabetes, facilitating a better understanding of disease development on the single cell scale. In situ monitoring is not a panacea; many cell environments would require labeling of cells to track their position and allow reliable, repeated sampling, particularly for systems containing millions to billions of cells.

An alternative route for increasing the chemical information detected within a single cell is to combine two or more techniques for multimodal analysis of individual cells. Multiple orthogonal approaches can increase the chemical information gained from one sample by merging the assets of each method. While simple in concept, multimodal analysis requires attention to details as each analytical technique has specific sample preparation requirements, substrates and conditions that may be incompatible with the other sampling protocols, and of course, the approaches must leave sufficient sample behind for the next measurement. In 2011, the Zenobi group[117] developed a method for coupling laser desorption/ionization MS with

Raman and fluorescence imaging to analyze carotene and phospholipids within individual algal cells randomly seeded on a metal target. Furthermore, they coupled Raman and fluorescence microscopy with MALDI MS to study adenosine triphosphate, adenosine diphosphate, and beta- carotene within individual algal cells.[109] These studies demonstrate that not only is the combination of MS methods with optical approaches possible, the union can enhance the chemical coverage of a single cell. We hyphenated various analytical techniques to obtain more complete measurements on small samples. In 2013, we used immunocytochemistry to label neurons within the insect, Periplaneta americana, with follow-up peptide analysis using MALDI

MS.[118] Combining different toolsets, patch clamp electrophysiology was coupled to CE-MS in

2014 to study the relation between physiological activity of gamma-aminobutyric acid containing neurons within the rat thalamus in relation to neurochemical condition.[43] More

39 recently, we combined several MS approaches together such as MALDI MS and SIMS on the same cells. In 2017 an optically-guided liquid microjunction extraction probe was utilized to hyphenate MALDI MS with CE-MS for analysis of single α and β pancreatic islets cells.[87]

MALDI MS is used first to prescreen cells for peptides indicative of a specific cell type, while

CE-MS is used to analyze metabolite profiles of cells representative of a cell type. This combination can be further employed to study both the small metabolite and peptide heterogeneity possessed by different cell types. We expect multimodal measurements to remain a rapidly developing approach for increasing chemical coverage. By coupling single cell MS techniques with complementary targeted and untargeted techniques, such as single cell transcriptomics, spectroscopy and staining, even more information can be garnered.

Looking back to the days of Schleiden and Schawnn[1] when cell theory was first described, the scientific community has learned much about this fundamental unit of life.

Numerous technological advances have greatly expanded the information content that can be obtained from single cell analyses. Systems that have been explored include algae, plants, sea slugs and mammals, suggesting that single cell analysis has been effectively transforming many areas of biological science. Future advancements may allow scientists to probe more complex systems and glean information on some of the least understood emergent properties of organisms, such as memory within the brain. While challenging, probing the chemical profiles and activity of individual cells is most certainly a field worth pursuing.

2.7 Acknowledgements

The authors gratefully acknowledge support from the National Institutes of Health,

Award Number P30 DA018310 from the National Institute on Drug Abuse, and from the

National Institute of Mental Health, Award Number 1U01 MH109062. E.K.N. and T.J.C.

40 acknowledge support from the National Science Foundation Graduate Research Fellowship

Program and the Springborn Fellowship. T.J.C. received additional support through the

Training Program at Chemistry-Interface with Biology (T32 GM070421).

41

2.8 Figures

Figure 2.1 Analytical techniques for single cell analysis discussed in the review. (Top)

Fractionation techniques include capillary electrophoresis combined with different detection methods. (Bottom) Cells are directly probed for their contents using a variety of ionization methods that have a range of capabilities and performance specifications.

42

Figure 2.2 Several examples of single cell CE-MS approaches highlighting the chemical information obtained. A) Comparison between CE electropherograms of the V1R cell dissected and probed twice consecutively. Similar metabolites were detected in either approach. B)

Significant changes in GABA and aspartate were detected between the progenitor cell and daughter cells, while creatine was not determined to be significantly different. Patch clamp profiles of two different neurons. A and B are adapted from Onjiko et al., Anal. Chem. 2017, 89,

7069–7076. Copyright 2017 American Chemical Society. C) Ventral basal thalamocortical and

D) thalamic reticular nucleus, as well as metabolic CE electropherograms of each cell. The authors detected ornithine, GABA, glycine, serine, tryptophan, glutamine, glutamate, tyrosine, and proline. C and D are adapted from Aerts et al., Anal Chem. 2014, 86, 3203–3208. Copyright

2014 American Chemical Society.

43

Figure 2.3 A series of cellular chemical images demonstrating the information that can be gained from SIMS. A) Anaerobic oxidation of methane of consortia of anaerobic methanotrophic archaea (green) paired with Deltaproteobacteria (pink) identified by fluorescence in situ hybridization (FISH) (left panels) and NanoSIMS imaging (middle panels). Single cell activities

(right panels) are measured as 15N atom percentages for regions of interest representing the

FISH-identified archaea and bacteria in each consortium. Lighter shaded cells are more enriched

15 15 + in N, which corresponds with higher levels of anabolic activity and NH4 assimilation. Scale bars, 3 µm. Adapted from McGlynn et al., Nature 2015, 526, 531–535. Copyright 2015 Nature

Publishing Group. B) Multidimensional, composite image obtained from multiplexed ion beam imaging (MIBI) illustrating quantitatively protein expression and colocalization of E-cadherin

(green), actin (red) and vimentin (blue), qualitatively categorization of cell nuclei into two 44

Figure 2.3 (cont.) subpopulations: ERα+PR+Ki-67+ (yellow) and ERα+PR+(aqua). Scale bars, 25

µm. Adapted from Angelo et al., Nat. Med. 2014, 20, 436–442. Copyright 2014 Nature

Publishing Group.

45

Figure 2.4 SIMS imaging can be used to determine the localization of a variety of different chemical classes within single cells. For instance, A) pooled signals of amino acid fragment ions are represented in red, B) those of phospholipids in green, and C) substrate-derived secondary ions are depicted in blue. D–F) Red–green–blue color overlay for correlation analysis of a confluent monolayer of normal rat kidney cells. Both horizontal xy and vertical xz sections are shown. Adapted from Breitenstein et al., Angew. Chem. Int. Ed. 2007, 46, 5332–5335.

Copyright 2007 Wiley InterScience. 46

Figure 2.5 High spatial resolution MALDI-MSI can be also be used to determine subcellular localization of chemical classes that are often different than those obtaining with SIMS imaging.

A, C and E) Optical images of P. caudatum before MALDI-MSI analysis. B, D and F) MSI

+ + images (RGB mode) of different lipids (i.e. [DG(31:0) + NH4] , [PS(32:1) + H - H2O] ,

+ + [DG(38:1) + NH4] ) or small peptides (e.g. [(Lys3Ala) + H] ) present within the cell. MS images were acquired at 3-µm spatial resolution, showing discrete subcellular locations of the different lipids or small peptides. Some of the metabolites correlated well with structures such as cilia.

Scale bar: 100 µm. Figure adapted from Kompauer et al., Nat. Methods 2017, 14, 90–96.

Copyright 2017 Nature Publishing Group.

47

Figure 2.6 Two approaches are highlighted that provide high throughput single cell measurements via direct profiling MALDI MS. A) Schematic of the MAMS process, where cells are seeded within evenly spaced wells for MS analysis, and of B) optically-guided MS analyses, where cells are randomly seeded on a glass slide. C) MAMS has facilitated multi- modal analysis of algal cells by both MALDI-MS and Raman spectroscopy. Adapted from Urban et al., Lab Chip 2010, 10, 3206–3209. Copyright 2010 The Royal Society of Chemistry. D)

Similarly, microMS has facilitated the analysis of the same pancreatic islet cell by both MALDI-

MS and CE-MS. This specific method allowed MALDI-MS to prescreen single islet cells for

CE-MS analysis of ß-cells. Adapted from Comi et al., Anal. Chem. 2017, 89, 7765–7772.

Copyright 2017 American Chemical Society.

48

2.9 References

(1) T. Schwann, Microscopic Investigations on the Accordance in theStructure and Growth

of Plants and Animals, English ed., The Sydenham Society: London, Berlin, 1839.

(2) H. P. Schmiedebach, Z. Arztl Fortbild (Jena) 1990, 84, 889-894.

(3) R. L. K. Virchow, Cellular Pathology, JohnChurchill, London, 1859.

(4) M. Schultz, Emerg. Infect. Dis. 2008, 14, 1480-1481.

(5) a) M. Crow, A. Paul, S. Ballouz, Z. J. Huang, J. Gillis, Genome Biol. 2016, 17, 101; b) K.

Hodne, F. A. Weltzien, Int. J. Mol. Sci. 2015, 16, 26832-26849; c) E. Shapiro, T.

Biezuner, S. Linnarsson, Nat. Rev. Genetics 2013, 14, 618-630.

(6) D. A. Haber, V. E. Velculescu, Cancer Discov. 2014, 4, 650-661.

(7) X. Chen, J. C. Love, N. E. Navin, L. Pachter, M. J. T. Stubbington, V. Svensson, J. V.

Sweedler, S. A. Teichmann, Nat. Biotechnol. 2016, 34, 1111-1118.

(8) T. J. Comi, T. D. Do, S. S. Rubakhin, J. V. Sweedler, J. Am. Chem. Soc. 2017, 139, 3920-

3929.

(9) R. T. Kennedy, J. W. Jorgenson, Anal. Chem. 1989, 61, 436-441.

(10) R. M. Wightman, J. A. Jankowski, R. T. Kennedy, K. T. Kawagoe, T. J. Schroeder, D. J.

Leszczyszyn, J. A. Near, E. J. Diliberto, Jr., O. H. Viveros, Proc. Natl. Acad. Sci. U. S. A.

1991, 88, 10754-10758.

(11) J. B. Chien, R. A. Wallingford, A. G. Ewing, J. Neurochem. 1990, 54, 633-638.

(12) S. A. Mackler, B. P. Brooks, J. H. Eberwine, Neuron 1992, 9, 539-548.

(13) R. Zenobi, Science 2013, 342, 1243259.

(14) C. Cecala, J. V. Sweedler, Analyst 2012, 137, 2922-2929.

(15) K. Chughtai, R. M. A. Heeren, Chem. Rev. 2010, 110, 3237-3277.

49

(16) a) X. Ye, F. Xie, E. V. Romanova, S. S. Rubakhin, J. V. Sweedler, ACS Chem. Neurosci.

2010, 1, 182-193; b) P. D. Floyd, L. Li, S. S. Rubakhin, J. V. Sweedler, C. C. Horn, I.

Kupfermann, V. Y. Alexeeva, T. A. Ellis, N. C. Dembrow, K. R. Weiss, F. S. Vilim, J

Neurosci 1999, 19, 7732-7741; c) H. Miao, S. S. Rubakhin, C. R. Scanlan, L. Wang, J. V.

Sweedler, J. Neurochem. 2006, 97, 595-606; d) L. L. Moroz, J. R. Edwards, S. V.

Puthanveettil, A. B. Kohn, T. Ha, A. Heyland, B. Knudsen, A. Sahni, F. Yu, L. Liu, S.

Jezzini, P. Lovell, W. Iannucculli, M. Chen, T. Nguyen, H. Sheng, R. Shaw, S.

Kalachikov, Y. V. Panchin, W. Farmerie, J. J. Russo, J. Ju, E. R. Kandel, Cell 2006, 127,

1453-1467.

(17) S. Hosic, S. K. Murthy, A. N. Koppes, Anal. Chem. 2016, 88, 354-380.

(18) M. E. Johnson, J. P. Landers, Electrophoresis 2004, 25, 3513-3527.

(19) R. A. Wallingford, A. G. Ewing, Anal. Chem. 1988, 60, 1972-1975.

(20) A. G. Ewing, J. M. Mesaros, P. F. Gavin, Anal. Chem. 1994, 66, A527-A537.

(21) L. Sun, K. M. Dubiak, E. H. Peuchen, Z. Zhang, G. Zhu, P. W. Huber, N. J. Dovichi,

Anal. Chem. 2016, 88, 6653-6657.

(22) L. A. Woods, T. P. Roddy, A. G. Ewing, Electrophoresis 2004, 25, 1181-1187.

(23) K. D. Altria, Capillary Electrophoresis Guidebook 1996, 3-13.

(24) P. Nemes, A. M. Knolhoff, S. S. Rubakhin, J. V. Sweedler, Anal. Chem. 2011, 83, 6810-

6817.

(25) C. E. Sims, G. D. Meredith, T. B. Krasieva, M. W. Berns, B. J. Tromberg, N. L.

Allbritton, Anal. Chem. 1998, 70, 4570-4577.

(26) F. T. Han, Y. Wang, C. E. Sims, M. Bachman, R. S. Chang, G. P. Li, N. L. Allbritton,

Anal. Chem. 2003, 75, 3688-3696.

50

(27) a) N. R. Munce, J. Z. Li, P. R. Herman, L. Lilge, Anal. Chem. 2004, 76, 4983-4989; b) J.

S. Mellors, K. Jorabchi, L. M. Smith, J. M. Ramsey, Anal. Chem. 2010, 82, 967-973; c)

A. D. Hargis, J. P. Alarie, J. M. Ramsey, Electrophoresis 2011, 32, 3172-3179; d) A. J.

Dickinson, P. M. Armistead, N. L. Allbritton, Anal. Chem. 2013, 85, 4797-4804.

(28) B. Brügger, Annu. Rev. Biochem 2014, 83, 79-98.

(29) L. L. Moroz, R. L. Dahlgren, D. Boudko, J. V. Sweedler, P. Lovell, J. Inorg. Biochem.

2005, 99, 929-939.

(30) S. Hu, Z. Le, S. Krylov, N. J. Dovichi, Anal. Chem. 2003, 75, 3495-3501.

(31) H. Lochmann, A. Bazzanella, S. Kropsch, K. Bachmann, J. Chromatogr. A 2001, 917,

311-317.

(32) Q. Zhi, C. Xie, X. Huang, J. Ren, Anal. Chim. Acta 2007, 583, 217-222.

(33) a) R. M. Onjiko, E. P. Portero, S. A. Moody, P. Nemes, Anal. Chem. 2017, 89, 7069-

7076; b) R. M. Onjiko, D. O. Plotnick, S. A. Moody, P. Nemes, Anal. Methods 2017, 9,

4964-4970.

(34) X. Wang, Y. Ma, M. Zhao, M. Zhou, Y. Xiao, Z. Sun, L. Tong, J. Chromatogr. A 2016,

1469, 128-134.

(35) C. Lombard-Banek, S. Reddy, S. A. Moody, P. Nemes, Molecular & Cellular Proteomics

: MCP 2016, 15, 2756-2768.

(36) J.-X. Liu, J. T. Aerts, S. S. Rubakhin, X.-X. Zhang, J. V. Sweedler, Analyst 2014, 139,

5835-5842.

(37) A. V. Patel, T. Kawai, L. Wang, S. S. Rubakhin, J. V. Sweedler, Anal. Chem. 2017, 89,

12375-12382.

51

(38) D. Chen, F. Fan, X. Zhao, F. Xu, P. Chen, J. Wang, L. Ban, Z. Liu, X. Feng, Y. Zhang,

B.-F. Liu, Anal. Chem. 2016, 88, 2466-2471.

(39) D. T. Chiu, S. J. Lillard, R. H. Scheller, R. N. Zare, S. E. Rodriguez-Cruz, E. R.

Williams, O. Orwar, M. Sandberg, J. A. Lundqvist, Science 1998, 279, 1190-1193.

(40) H. N. Li, C. E. Sims, H. Y. Wu, N. L. Allbritton, Anal. Chem. 2001, 73, 4625-4631.

(41) N. Gunasekera, K. J. Olson, K. Musier-Forsyth, E. A. Arriaga, Anal. Chem. 2004, 76,

655-662.

(42) R. D. Johnson, M. Navratil, B. G. Poe, G. H. Xiong, K. J. Olson, H. Ahmadzadeh, D.

Andreyev, C. F. Duffy, E. A. Arriaga, Anal. Bioanal. Chem. 2007, 387, 107-118.

(43) J. T. Aerts, K. R. Louis, S. R. Crandall, G. Govindaiah, C. L. Cox, J. V. Sweedler, Anal.

Chem. 2014, 86, 3203-3208.

(44) a) A. Benninghoven, Angew. Chem. Int. Ed. 1994, 33, 1023-1043; b) J. L. Guerquin-

Kern, T. D. Wu, C. Quintana, A. Croisy, Biochim. Biophys. Acta 2005, 1724, 228-238; c)

J. S. Fletcher, Analyst 2009, 134, 2204-2215.

(45) a) S. G. Boxer, M. L. Kraft, P. K. Weber, Annu. Rev. Biophys. 2009, 38, 53-74; b) L. J.

Gamble, C. R. Anderton, Micros. Today 2016, 24, 24-31.

(46) a) R. Levi-Setti, J. M. Chabala, K. Gavrilov, R. Espinosa, 3rd, M. M. Le Beau, Microsc.

Res. Tech. 1997, 36, 301-312; b) R.Levi-Setti, K.L.Gavrilov, P.L.Strissel, R.Strick, Appl.

Surf. Sci. 2004, 231-232, 479-484; c) W. Romer, T.-D. Wu, P. Duchambon, M. Amessou,

D. Carrez, L. Johannes, J.-L. Guerquin-Kern, Appl. Surf. Sci. 2006, 252, 6925-6930.

(47) a) X. Zha, W. A. Ausserer, G. H. Morrison, Cancer Res. 1992, 52, 5219-5222; b) D. R.

Lorey, G. H. Morrison, S. Chandra, Anal. Chem. 2001, 73, 3947-3953.

(48) S. Chandra, D. R. L. II, Int. J. Mass spectrom. 2007, 260.

52

(49) N. Musat, R. Foster, T. Vagner, B. Adam, M. M. Kuypers, FEMS Microbiol. Rev. 2012,

36, 486-511.

(50) C. P. Lechene, Y. Luyten, G. McMahon, D. L. Distel, Science 2007, 318, 570-570.

(51) R. Popa, P. K. Weber, J. Pett-Ridge, J. A. Finzi, S. J. Fallon, I. D. Hutcheon, K. H.

Nealson, D. G. Capone, ISME J. 2007, 1, 354-360.

(52) N. Musat, H. Halm, B. Winterholler, P. Hoppe, S. Peduzzi, F. Hillion, F. Horreard, R.

Amann, B. B. Jorgensen, M. M. Kuypers, Proc. Natl. Acad. Sci. U. S. A. 2008, 105,

17861-17866.

(53) a) S. E. McGlynn, G. L. Chadwick, C. P. Kempes, V. J. Orphan, Nature 2015, 526, 531-

535; b) V. J. Orphan, C. H. House, K. U. Hinrichs, K. D. McKeegan, E. F. DeLong,

Science 2001, 293, 484-487; c) A. E. Dekas, R. S. Poretsky, V. J. Orphan, Science 2009,

326, 422-426; d) D. A. Fike, C. L. Gammon, W. Ziebis, V. J. Orphan, ISME J. 2008, 2,

749-759.

(54) S. A. Eichorst, F. Strasser, T. Woyke, A. Schintlmeister, M. Wagner, D. Woebken, FEMS

Microbiol. Ecol. 2015, 91.

(55) a) S. K. Saka, A. Vogts, K. Krohnert, F. Hillion, S. O. Rizzoli, J. T. Wessels, Nat.

Commun. 2014, 5, 3664; b) N. T. N. Phan, X. Li, A. G. Ewing, Nat. Rev. Chem. 2017, 1;

c) A. A. Legin, A. Schintlmeister, M. A. Jakupec, M. Galanski, I. Lichtscheidl, M.

Wagner, B. K. Keppler, Chem. Sci. 2014, 5, 3135-3143.

(56) M. H. Spitzer, G. P. Nolan, Cell 2016, 165, 780-791.

(57) M. Angelo, S. C. Bendall, R. Finck, M. B. Hale, C. Hitzman, A. D. Borowsky, R. M.

Levenson, J. B. Lowe, S. D. Liu, S. Zhao, Y. Natkunam, G. P. Nolan, Nat. Med. 2014,

20, 436-442.

53

(58) F. R. Moss, 3rd, S. G. Boxer, J. Am. Chem. Soc. 2016, 138, 16737-16744.

(59) S. G. Ostrowski, C. T. Van Bell, N. Winograd, A. G. Ewing, Science 2004, 305, 71-73.

(60) E. B. Monroe, J. C. Jurchen, J. Lee, S. S. Rubakhin, J. V. Sweedler, J. Am. Chem. Soc.

2005, 127, 12152-12153.

(61) K. R. Tucker, Z. Li, S. S. Rubakhin, J. V. Sweedler, J. Am. Soc. Mass Spectrom. 2012,

23, 1931-1938.

(62) a) S. C. C. Wong, R. Hill, P. Blenkinsopp, N. P. Lockyer, D. E. Weibel, J. C. Vickerman,

Appl. Surf. Sci. 2003, 219, 203-204; b) D. Weibel, S. Wong, N. Lockyer, P. Blenkinsopp,

R. Hill, J. C. Vickerman, Anal. Chem. 2003, 75, 1754-1764.

(63) J. S. Fletcher, S. Rabbani, A. Henderson, P. Blenkinsopp, S. P. Thompson, N. P.

Lockyer, J. C. Vickerman, Anal. Chem. 2008, 80, 9058-9064.

(64) M. K. Passarelli, A. G. Ewing, N. Winograd, Anal. Chem. 2013, 85, 2231-2238.

(65) E. J. Lanni, S. J. Dunham, P. Nemes, S. S. Rubakhin, J. V. Sweedler, J. Am. Soc. Mass

Spectrom. 2014, 25, 1897-1907.

(66) J. S. Fletcher, J. C. Vickerman, Anal. Chem. 2013, 85, 610-639.

(67) D. Breitenstein, C. E. Rommel, R. Mollers, J. Wegener, B. Hagenhoff, Angew. Chem. Int.

Ed. 2007, 46, 5332-5335.

(68) H. Nygren, B. Hagenhoff, P. Malmberg, M. Nilsson, K. Richter, Micros. Res. Tech. 2007,

70, 969-974.

(69) M. A. Robinson, D. J. Graham, D. G. Castner, Anal. Chem. 2012, 84, 4880-4885.

(70) J. Brison, M. A. Robinson, D. S. W. Benoit, S. Muramoto, P. S. Stayton, D. G. Castner,

Anal. Chem. 2013, 85, 10869-10877.

54

(71) J. S. Fletcher, N. P. Lockyer, S. Vaidyanathan, J. C. Vickerman, Anal. Chem. 2007, 79,

2199-2206.

(72) J. S. Fletcher, S. Rabbani, A. Henderson, N. P. Lockyer, J. C. Vickerman, Rapid

Commun. Mass Spectrom. 2011, 25, 925-932.

(73) H. Tian, D. A. Six, T. Krucker, J. A. Leeds, N. Winograd, Anal. Chem. 2017, 89, 5050-

5057.

(74) a) R. A. Kruse, S. S. Rubakhin, E. V. Romanova, P. W. Bohn, J. V. Sweedler, J. Mass

Spectrom. 2001, 36, 1317-1322; b) Y. Schober, S. Guenther, B. Spengler, A. Rompp,

Anal. Chem. 2012, 84, 6293-6297.

(75) A. J. Ibanez, S. R. Fagerer, A. M. Schmidt, P. L. Urban, K. Jefimovs, P. Geiger, R.

Dechant, M. Heinemann, R. Zenobi, Proc. Natl. Acad. Sci. U.S.A. 2013, 110, 8790-8794.

(76) A. El-Aneed, A. Cohen, J. Banoub, App. Spec. Rev. 2009, 44, 210-230.

(77) R. W. Garden, L. L. Moroz, T. P. Moroz, S. A. Shippy, J. V. Sweedler, J. Mass

Spectrom. 1996, 31, 1126-1130.

(78) J. H. Yang, R. M. Caprioli, Anal. Chem. 2011, 83, 5728-5734.

(79) R. L. Caldwell, R. M. Caprioli, Mol. Cell. Proteomics 2005, 4, 394-401.

(80) D. L. Baluya, T. J. Garrett, R. A. Yost, Anal. Chem. 2007, 79, 6862-6867.

(81) R. M. Caprioli, T. B. Farmer, J. Gile, Anal. Chem. 1997, 69, 4751-4760.

(82) V. Labas, A. P. Teixeira-Gomes, L. Bouguereau, A. Gargaros, L. Spina, A. Marestaing,

S. Uzbekova, J. Proteomics 2017.

(83) a) S. S. Rubakhin, J. D. Churchill, W. T. Greenough, J. V. Sweedler, Anal. Chem. 2006,

78, 7267-7272; b) C. R. Jimenez, P. A. Vanveelen, K. W. Li, W. C. Wildering, W. P. M.

Geraerts, U. R. Tjaden, J. Vandergreef, J. Neurochem. 1994, 62, 404-407.

55

(84) a) P. L. Urban, K. Jefimovs, A. Amantonico, S. R. Fagerer, T. Schmid, S. Madler, J.

Puigmarti-Luis, N. Goedecke, R. Zenobi, Lab on a Chip 2010, 10, 3206-3209; b) M. A.

Hossen, Y. Nagata, M. Waki, Y. Ide, S. Takei, H. Fukano, G. A. Romero-Perez, S.

Tajima, I. Yao, K. Ohnishi, M. Setou, Anal. Bioanal. Chem. 2015, 407, 5273-5280.

(85) a) A. Amantonico, J. Y. Oh, J. Sobek, M. Heinemann, R. Zenobi, Angew. Chem. Int. Ed.

2008, 47, 5382-5385; b) A. Amantonico, P. L. Urban, S. R. Fagerer, R. M. Balabin, R.

Zenobi, Anal. Chem. 2010, 82, 7394-7400.

(86) J. S. Page, J. V. Sweedler, Anal. Chem. 2002, 74, 6200-6204.

(87) T. J. Comi, M. A. Makurath, M. C. Philip, S. S. Rubakhin, J. V. Sweedler, Anal. Chem.

2017, 89, 7765-7772.

(88) A. Rompp, S. Guenther, Y. Schober, O. Schulz, Z. Takats, W. Kummer, B. Spengler,

Angew. Chem. Int. Ed. 2010, 49, 3834-3838.

(89) A. Zavalin, E. M. Todd, P. D. Rawhouser, J. Yang, J. L. Norris, R. M. Caprioli, J. Mass

Spectrom. 2012, 47, 1473-1481.

(90) M. Kompauer, S. Heiles, B. Spengler, Nat. Methods 2017, 14, 90-96.

(91) J. K. Lee, E. T. Jansson, H. G. Nam, R. N. Zare, Anal. Chem. 2016, 88, 5453-5461.

(92) R. Yin, J. Kyle, K. Burnum-Johnson, K. J. Bloodsworth, L. Sussel, C. Ansong, J. Laskin,

Anal. Chem. 2018, 90, 6548-6555.

(93) N. Pan, W. Rao, N. R. Kothapalli, R. M. Liu, A. W. G. Burgett, Z. B. Yang, Anal. Chem.

2014, 86, 9376-9380.

(94) W. Rao, N. Pan, Z. B. Yang, J. Am. Soc. Mass Spectrom. 2015, 26, 986-993.

(95) P. A. Vanveelen, C. R. Jimenez, K. W. Li, W. C. Wildering, W. P. M. Geraerts, U. R.

Tjaden, J. Vandergreef, Org. Mass Spectrom. 1993, 28, 1542-1546.

56

(96) a) C. R. Jimenez, K. W. Li, K. Dreisewerd, S. Spijker, R. Kingston, R. H. Bateman, A. L.

Burlingame, A. B. Smit, J. van Minnen, W. P. M. Geraerts, Biochemistry 1998, 37, 2070-

2076; b) E. Kellett, S. J. Perry, N. Santama, B. M. Worster, P. R. Benjamin, J. F. Burke,

J. Neurosci. 1996, 16, 4949-4957; c) K. W. Li, R. Kingston, K. Dreisewerd, C. R.

Jimenez, R. C. vanderSchors, R. H. Bateman, W. P. M. Geraerts, Anal. Chem. 1997, 69,

563-565; d) S. J. Perry, A. C. Dobbins, M. G. Schofield, M. R. Piper, P. R. Benjamin,

Eur. J. Neurosci. 1999, 11, 655-662; e) R. W. Garden, T. P. Moroz, J. M. Gleeson, P. D.

Floyd, L. J. Li, S. S. Rubakhin, J. V. Sweedler, J. Neurochem. 1999, 72, 676-681; f) F. J.

van Strien, S. Jespersen, J. van der Greef, B. G. Jenks, E. W. Roubos, FEBS Lett. 1996,

379, 165-170.

(97) a) L. J. Li, R. W. Garden, E. V. Romanova, J. V. Sweedler, Anal. Chem. 1999, 71, 5451-

5458; b) L. J. Li, R. W. Garden, J. V. Sweedler, Trends Biotechnol. 2000, 18, 151-160; c)

K. J. Boggio, E. Obasuyi, K. Sugino, S. B. Nelson, N. Y. Agar, J. N. Agar, Expert Rev.

Proteomics 2011, 8, 591-604; d) L. J. Li, E. V. Romanova, S. S. Rubakhin, V. Alexeeva,

K. R. Weiss, F. S. Vilim, J. V. Sweedler, Anal. Chem. 2000, 72, 3867-3874; e) J. S. Page,

S. S. Rubakhin, J. V. Sweedler, Anal. Chem. 2002, 74, 497-503.

(98) E. Sugiyama, I. Yao, M. Setou, Visualization of local phosphatidylcholine synthesis

within hippocampal neurons using a compartmentalized culture system and imaging

mass spectrometry, Vol. 495, 2017.

(99) L. C. Chen, K. Nishidate, Y. Saito, K. Mori, D. Asakawa, S. Takeda, T. Kubota, N.

Terada, Y. Hashimoto, H. Hori, K. Hiraoka, Rapid Commun. Mass Spectrom. 2008, 22,

2366-2374.

57

(100) X. Y. Gong, Y. Y. Zhao, S. Q. Cai, S. J. Fu, C. D. Yang, S. C. Zhang, X. R. Zhang, Anal.

Chem. 2014, 86, 3809-3816.

(101) L. Zhang, C. J. Sevinsky, B. M. Davis, A. Vertes, Anal. Chem. 2018, 90, 4626-4634.

(102) H. Zhu, G. Zou, N. Wang, M. Zhuang, W. Xiong, G. Huang, Proceedings of the National

Academy of Sciences 2017, 114, 2586-2591.

(103) H. Mizuno, N. Tsuyama, T. Harada, T. Masujima, J. Mass Spectrom. 2008, 43, 1692-

1700.

(104) a) T. Fujii, S. Matsuda, M. L. Tejedor, T. Esaki, I. Sakane, H. Mizuno, N. Tsuyama, T.

Masujima, Nat. Protoc. 2015, 10, 1445-1456; b) M. L. Tejedor, H. Mizuno, N. Tsuyama,

T. Harada, T. Masujima, Anal. Chem. 2012, 84, 5221-5228.

(105) G. Li, S. Yuan, S. Zheng, Y. Liu, G. Huang, Anal. Chem. 2018, 90, 3409-3415.

(106) R. Yin, V. Prabhakaran, J. Laskin, Anal. Chem. 2018, 90, 7937-7945.

(107) T. H. Ong, D. J. Kissick, E. T. Jansson, T. J. Comi, E. V. Romanova, S. S. Rubakhin, J.

V. Sweedler, Anal. Chem. 2015, 87, 7036-7042.

(108) M. Pabst, S. R. Fagerer, R. Kohling, K. Eyer, J. Krismer, K. Jefimovs, A. J. Ibanez, R.

Zenobi, J. Am. Soc. Mass Spectrom. 2014, 25, 1083-1086.

(109) S. R. Fagerer, T. Schmid, A. J. Ibanez, M. Pabst, R. Steinhoff, K. Jefimovs, P. L. Urban,

R. Zenobi, Analyst 2013, 138, 6732-6736.

(110) E. T. Jansson, T. J. Comi, S. S. Rubakhin, J. V. Sweedler, ACS Chem. Biol. 2016, 11,

2588-2595.

(111) T. J. Comi, E. K. Neumann, T. D. Do, J. V. Sweedler, J. Am. Soc. Mass Spectrom. 2017,

28, 1919-1928.

58

(112) T. D. Do, J. F. Ellis, E. K. Neumann, T. J. Comi, E. G. Tillmaand, A. E. Lenhart, S. S.

Rubakhin, J. V. Sweedler, Chemphyschem : a European journal of chemical physics and

physical chemistry, 0.

(113) B. Yang, N. H. Patterson, T. Tsui, R. M. Caprioli, J. L. Norris, J. Am. Soc. Mass.

Spectrom. 2018, 29, 1012-1020.

(114) T. D. Do, T. J. Comi, S. J. Dunham, S. S. Rubakhin, J. V. Sweedler, Anal. Chem. 2017,

89, 3078-3086.

(115) P. Nemes, S. S. Rubakhin, J. T. Aerts, J. V. Sweedler, Nat. Protoc. 2013, 8, 783-799.

(116) T. L. Colliver, C. L. Brummel, M. L. Pacholski, F. D. Swanek, A. G. Ewing, N.

Winograd, Anal. Chem. 1997, 69, 2225-2231.

(117) P. L. Urban, T. Schmid, A. Amantonico, R. Zenobi, Anal. Chem. 2011, 83, 1843-1849.

(118) S. Neupert, S. S. Rubakhin, J. V. Sweedler, Chem. Biol. 2012, 19, 1010-1019.

59

CHAPTER 3

MICROMS: A PYTHON PLATFORM FOR IMAGE-GUIDED MASS SPECTROMETRY

PROFILING

Notes and Acknowledgements

This chapter is adapted from Troy J. Comi, Thanh D. Do, and Jonathan V. Sweedler, submitted for publication on March 28, 2017 and accepted June 7, 2017 in Journal of the

American Society for Mass Spectrometry, DOI: 10.1007/s13361-017-1704-1. T.J. Comi created the microMS code and wrote the majority of the manuscript, T. D. Do tested the code and performed the SIMS and MALDI MS comparisons. We acknowledge Prof. Stanislav Rubakhin for sample preparation assistance and Dr. Tong Si for useful discussions. The authors gratefully acknowledge support from the National Institutes of Health, Award Number P30 DA018310 from the National Institute on Drug Abuse, and from the National Institute of Mental Health,

Award Number 1U01 MH109062. E.K. Neumann and T.J. Comi acknowledge support from the

National Science Foundation Graduate Research Fellowship Program and the Springborn

Fellowship. T.J. Comi received additional support through the Training Program at Chemistry-

Interface with Biology (T32 GM070421).

3.1 Introduction

Image-guided mass spectrometry (MS) provides a link between the spatial dimensions in a digital image and the physical location of a sample within a microprobe system. MS imaging

(MSI) is a subset of image guided chemical sampling that frequently utilizes regularly spaced acquisition positions overlaid on an optical scan to recreate the spatial distribution of analytes within a sample. However, traditional MSI is low throughput and less sensitive than targeted profiling when the target objects (e.g. biological cells and bacterial colonies) are widely

60 dispersed or smaller than the microprobe size. In the past decades, single cell analysis with MS has attracted great interest due to its sensitivity and ability to handle volume-limited samples.[1-6]

Many classes of biomolecules within individual cells are detectible with a variety of MS probes, facilitating new discoveries of single cell heterogeneity and a better understanding of the relationship between chemical contents and cellular functions. When MSI is applied to tissue sections,[7-10] the resolution to differentiate neighboring cells requires sampling each cell multiple times, effectively splitting available analytes among pixels. Due to difficulties in sample preparation and stringent instrument requirements, MSI at or below single cell resolution is far from routine in most laboratories. In the case of dispersed cells,[11-12] traditional MS imaging is not an optimal approach as most of the measurement time is spent characterizing the space between the cells. The limitations of MSI for high throughput analysis of single cells have led to the development of new methods to locate or deposit cells.

Recently, high throughput approaches to single cell MS have driven analyses of dissociated single cells which are either chemically labeled[13] or coordinate registered. MS profiling of adhered cells provides advantages in data fusion by simplifying data processing and allowing sequential analysis of the same cell. Microarrays for mass spectrometry[14-16] (MAMS) have demonstrated such capabilities by combining Raman microspectroscopy with MS.[17] As an alternative to MAMS, cells may be randomly seeded on a substrate, greatly relaxing fabrication requirements at the expense of a necessarily gentle sample extraction. Ong et al. presented such an approach, by locating single cells on an indium tin oxide (ITO)-coated glass slide based on their position in a whole-slide fluorescence microscopy image.[18] A challenge with this initial report was the complex scheme for generating custom geometry files, which required manual interaction through several disjointed pieces of software. To facilitate broader adoption of

61 optically-guided single cell profiling, we sought to streamline the process of directing MS acquisition with whole-slide microscopy images. As reported by Jansson et al., the first iteration utilized a point-based similarity registration scheme, which improved target localization accuracy over the previously reported piece-wise linear transform.[19] User interaction was also simplified, allowing fluid interaction with microscope images through a graphical user interface

(GUI). All functions required to begin acquiring single cell mass spectra on a Bruker ultrafleXtreme instrument were contained in the single piece of software.

Here we present the first version of microMS to support microscopy-guided MS for a variety of image files and mass spectrometers. The software architecture permits new microprobe instruments to be supported with minor modifications to the source code. Virtually any spatially restricted sampling probe capable of precisely recording and moving to a given location can perform such profiling.

First, the unique features of microMS are described along with the necessary modifications to expand device support for both commercial and customized instruments. We then illustrate an example of using microMS on three MS systems for off-line, targeted profiling of single cells from the mammalian nervous system.

3.2 Materials and Methods

Software microMS is written in python v3.5. In addition to base components, microMS requires the matplotlib, PyQt5, numpy, scipy, openslide, skimage, pyserial packages. Installation instructions, usage details and most recent source code may be found at github.com/troycomi/microms.

The program structure is modeled in Supporting Information Figure S1, with a partial representation in Figure 1. The main GUI class is composed of two widgets in the GUICanvas

62 package for displaying a microscope image or population-level statistics as a histogram. Each widget interacts with a microMSModel object, which represents a single microscopy experiment

(as a blobList and slideWrapper) and mass spectrometer (as a coordinateMapper).

Targets in microMS are represented as objects called “blobs”, to generalize a biological cell as any object formed by a group of high intensity pixels. In Figure 3.2, there are three blobs; each with a unique Cartesian (x, y) location on the image, a corresponding effective radius, and circularity. The circularity is a ratio of the blob area to its perimeter squared, scaled between 0 and 1, with 1 being a perfect circle, i.e. blobs 1 and 3 are single cells whereas blob 2 is not. A collection of blobs is stored in a blobList object, which also implements methods to query and filter a population of targets.

A slideWrapper provides an object for interacting with a set of microscopy images representing brightfield and multiple fluorescence channel images. The current field of view is maintained to simplify controller interaction with the image. The ImageUtilities package also contains modules for cell finding, patterning target positions, and optimizing travel paths.

Object models for MS instruments are contained in the coordinateMappers package, as shown in the model in Figure 3.1. The coordinateMapper is an abstract base class providing an interface which the GUI software utilizes to interact with different instrument systems. The core functionality of the mapper is to align pixel positions with physical coordinates and provide a means to translate target positions on an image to instrument-specific directions. The design of the software architecture simplifies the addition of new instruments. Integration of ambient ionization methods, including the single-probe[20-21] or nanoDESI[22] are enticing candidates as they have demonstrated single cell sensitivities in imaging and profiling applications. Currently, four concrete implementations are supplied in the CoordianteMappers package: a Bruker

63

UltrafleXtreme, a Bruker SolariX, the AB Sciex oMALDI sample stage attached to a custom

+ hybrid MALDI/C60 -SIMS, and a lab-built 3-axis liquid microjunction probe. Details about the implementations will be discussed in the next section. Demonstrations for the addition of new instruments to microMS may be found in the user manual packaged with the source code.

Single cell dissociation.

All chemicals were purchased from Sigma Aldrich (St. Louis, MO) and used without further purification. Two, 2-2.5-month-old male Sprague Dawley outbred rats (Rattus norvegicus) (www.envigo.com) were housed on a 12-h light cycle and fed ad libitum. Animal euthanasia was performed in accordance with the appropriate institutional animal care guidelines

(the Illinois Institutional Animal Care and Use Committee), and in full compliance with federal guidelines for the humane care and treatment of animals. Dissected cerebellum and suprachiasmatic nucleus (SCN) tissues were incubated in a solution of 1% Hoechst 33342 in oxygenated modified Gey’s balanced salt solution (mGBSS) for 30 minutes at 37°C. The mGBSS solution was removed and the tissues were incubated in an oxygenated solution of 6 units of papain, 1 mM L-cysteine, and 0.5 mM ethylenediaminetetraacetic acid for 80 minutes at

37°C. Tissue was then mechanically dissociated in mGBSS with 0.04% paraformaldehyde. A solution of 80% glycerol in mGBSS was added to a final concentration of 40% glycerol. The cell suspension was then transferred onto ITO-coated glass slides (Delta Technologies, Loveland,

CO) with at least 12 fiducial marks etched by a diamond-tipped pen.

Microscopy Imaging.

Brightfield and fluorescence images were acquired on a Zeiss Axio Imager M2 (Zeiss,

Jena, Germany) equipped with an Ab cam Icc5 camera, X-CITE Series 120 Q mercury lamp

(Lumen Dynamics, Mississauga, Canada), and a HAL 100 halogen illuminator (Zeiss, Jena,

64

Germany). The 31000v2 DAPI filter set was used for fluorescence excitation. The images were acquired in mosaic mode with a 10x objective and 13% overlap. Images were processed and exported as tiff files using ZEN software version 2 blue edition (Zeiss, Jena, Germany).

Sample Preparation.

Slides were coated with 50 mg/mL 2,5-dihydroxybenzoic acid (DHB) dissolved in 1:1

(v/v) LC-grade ethanol:water with 0.1% trifluoroacetic acid with an automatic sprayer described

[19,23] previously. The matrix solution was supplied at 10 mL/hr and nebulized with N2 gas at 50 psi over 100 passes. Samples were affixed to a rotating plate with the nebulizer positioned 1.5 cm above the samples, resulting in a MALDI matrix thickness of ~0.1-0.2 mg/cm2.

Instrument Parameters.

Single cell analysis was performed on three instruments. The UltrafleXtreme mass spectrometer (Bruker Daltonics, Billerica, MA) was set with a mass window of m/z 400-3000.

The “Ultra” (~100 mm footprint) laser setting was used with 300 laser shots at 1000 Hz for each cell to generate a matrix assisted laser desorption/ionization (MALDI)-TOF spectrum at each cell. The second instrument is a 7T SolariX FT-ICR mass spectrometer (Bruker Daltonics,

Billerica, MA), operated with a mass window of m/z 150-3000, yielding a 4 Mword time-domain transient. Spectra were calibrated to the phosphatidylcholine headgroup at m/z 184.07332.

Adsorption mode was used to effectively double the mass resolving power. Each MALDI spectrum was acquired with 20 laser shots at 1000 Hz and 60% laser energy. The laser setting

+ produced a ~100 mm footprint. The last instrument is a custom hybrid MALDI/C60 Q-TOF mass

[11] + spectrometer, described in detail elsewhere. The C60 ion beam was utilized for secondary ion mass spectrometry (SIMS), with the mass analyzer operated in positive mode with a mass range

65 of m/z 60-850. Correctly parsing the spectra requires additional instrument modifications and data analysis routines, described elsewhere.[24]

3.3 Results and Discussion

Instrument support in microMS.

Bruker instruments are discussed first as their MALDI sample stages and coordinate mappers are similar. The commonality is exploited by the brukerMapper abstract base class, which is a derived class of coordinateMapper. BrukerMapper implements methods for reading and writing xeo geometry files, which are required in Bruker software for automatic acquisition.

The brukerMapper class also defines an intermediate coordinate system between physical, motor coordinates and the fractional distances used in xeo files. The classes derived from brukerMapper require a limited set of concrete method implementations to be fully functional as many features are supported in the bases class. The simplest case is ultraflexMapper, for the

Bruker ultrafleXtreme instrument, which defines the required methods to parse user input.

The solarixMapper class for a Bruker SolariX instrument is similar to the ultraflexMapper class with three minor modifications: 1) the xeo files are limited to 400 positions, 2) an xlsx Excel file is also required for automatic acquisition, and 3) input coordinates are read directly from the system clipboard. The flexImagingSolarix object extends solarixMapper and overrides the saved file format for import in flexImaging software. These two instruments provide examples of supporting microscopy-guided MS on Bruker MALDI sources.

Instruments from other vendors inherit directly from the coordinateMapper. One example is the oMALDIMapper for interfacing with an AB Sciex oMALDI server. With this instrument, the sample positions are encoded in ptn pattern files which contain an x,y coordinate relative to the starting position with calibrated motor steps. Hence, in comparison to brukerMapper,

66 oMALDIMapper transforms motor coordinates to ptn coordinates instead of fractional distances.

To further simplify correlation of mass spectra to image coordinates, a corresponding text file is also exported with pixel positions of each target. As a final consideration, the sample stage was found to have significant motor slop upon changing direction. The motor slop is corrected before exporting the ptn file to ensure accurate targeting.

In the preceding examples, microscopy images are correlated with physical positions on a mass spectrometer sample stage to generate instrument-specific target coordinates. This off-line workflow is ideal for instruments lacking support for external control of the sample stage, as is usually the case. To demonstrate capabilities with on-line analysis and instrument control, additional interfaces were developed for controlling Zaber linear actuators. The zaberMapper class contains a simple implementation of the abstract base class coordinateMapper. It also has an instance variable connectedInstrument, which is used by microMS to interact with the sample stage. Another abstract base class, connectedInstrument specifies the method signatures necessary for a connected instrument. The concrete implementation provided is a system with three linear actuator stages, zaber3axis. This module inherits from zaberInterface, containing serial wrappers to simplify interaction with each stage, and implements the connectedInstrument interface. In addition to reading the current, physical position for coordinate registration, the user directs stage movement on the optical image or with key strokes. microMS functionalities.

General features of microMS include locating targets, filtering the target population, patterning each target, and coordinating the image with the physical location of a mass spectrometer stage. Only the last step is instrument specific. Users should refer to the User Guide in Supporting Information for a comprehensive illustration of these functionalities.

67

(a) Locating targets. Targets may be specified on the microscope image either by manually selecting locations or by performing automatic blob finding. In the former approach, targets are added via holding the “shift” key and left mouse clicking on the center of the feature of interest. This generates a blob of default radius and circularity of 1. A custom radius is specified by clicking and dragging from the circumference to the center of the feature.

In automatic blob finding, the search takes place over the entire image area unless a region of interest (ROI) is specified. ROIs are defined by clicking and dragging a rectangular area or drawing the region a vertex at a time. The blob finding algorithm thresholds the specified image color and then groups together pixels above that threshold, as shown in Figure 2B. Each group is then evaluated for its size and circularity. Putative blobs falling outside the user- specified parameters are discarded. Several features are available to assist with selection of suitable blob finding parameters. Different blob finding parameters are interactively tested on the current image field of view. Additionally, microMS reports pixel intensities, object size and circularity of positions selected with a middle mouse button click. Judicious selection of these parameters will find most cells while excluding imaging and background artifacts.

After the targets are located, their properties are stored as lists within microMS. Up to ten separate target lists are maintained and each list is displayed as a different color. New target sets generated by filtering or patterning are automatically stored in an empty list with the original target set left intact.

(b) Filtering targets. Frequently, it is beneficial to filter the target list, either to refine putative blobs or to stratify targets based on morphology. Basic filtering methods provided by microMS are selected through the menu bar which include ROI filtering and distance filtering.

Within a specific ROI, the blobs can be selectively removed or exclusively retained in a new

68 target list. ROI filtering is especially useful for removing targets which are near fiducials or potentially contaminated by substrate background. Distance filtering helps to ensure each MS target position will correspond to a unique object (e.g. a single cell). An appropriate value for distance filtering is chosen based on the microprobe size, target accuracy, and the desired number of samples per blob. For example, a 100 µm diameter probe on a system with 50 µm target accuracy would require distance filtering of at least 100 µm to minimize the chance of sampling a nearby blob. During distance filtering, any target with a neighbor closer than the specified value is removed from the blob list.

In addition to common filtering functions, microMS supports interactive examination of population-level statistics through the histogram window to partition the target list (Figure 3.3).

Metrics include blob size, circularity, nearest neighbor distance, and fluorescence intensity. Note that there is some redundancy between histogram filtering and blob finding. The overlap permits the selection of lenient blob finding parameters to exhaustively identify all putative cells which are then refined to the final target set. Such a scheme allows distance filtering to identify all possible contaminating objects and sub-classification based on size.

High and low pass filters of the targets may be defined on the histogram window. Targets falling within a filter range are dynamically displayed on the microscope image in the corresponding color (Figure 3.3B-C). Selecting a blob in the microscope image highlights its value on the histogram to assist with defining filter limits. High and low pass filters define new target lists that may be further refined by additional histogram operations. This function allows operations such as filtering a population based on size followed by selection of targets with a particular fluorescent stain. More routinely, the histogram provides a simple method to identify

69 and remove artifacts from blob finding. Figure 3.3C shows an example of isolating unresolved cells, which helps ensure data quality.

(c) Patterning targets. By default, microMS generates one acquisition target per blob.

Single target sampling is sufficient when the microprobe size is similar to or larger than the target object. However, when the object is larger than the probe, a single acquisition is insufficient to robustly sample heterogeneous objects. Alternatively, MSI of each blob can be acquired at each target location. To address advanced sampling requirements, microMS provides three sample patterning schemes.

The first option is a rectangular packed array of points centered on the target. Users select a raster spacing and number of layers to define the overall size of the image. Alternatively, the size is dynamically adjusted to the target radius to ensure complete sampling of heterogeneously sized populations. The resulting data is directly interpretable as an MS image with common MSI software.

Similar to rectangular packing is the hexagonal close packing pattern. With a circular desorption probe, the hexagonal packing provides denser sampling of the target. Users define the target separation, number of layers, and specify dynamic layering. While hexagonally packed data are more difficult to reconstruct into an image, averaging the spectra yields a representative spectrum for the blob.

Finally, the circular pattern generates targets around the circumference of each blob. In some cases, analyzing the center of a blob produces low sensitivity due to the morphology of the target or the biological nature of the samples. Instead, targets are placed immediately outside of a blob to acquire representative spectra. For circular patterning, the user defines a minimum target- to-target distance, maximum number of targets, and offset from the circumference. The actual

70 number of targets around a blob is determined by the blob size and the specified offset while maintaining the target-to-target distance above the minimum limit. Targets are then equally spaced around the blob. Averaging the resulting data provides a characteristic spectrum of the area directly surrounding each blob.

(d) Coordinate transformation to instrument systems. Once all targets are determined, the pixel positions must be translated into the physical coordinates of a mass spectrometer or similar platform. Image correlation in microMS is accomplished through a point-based similarity registration. In point-based registration, the target localization error scales inversely with the square root of the number of fiducial points. As such, while microMS supports arbitrary numbers of fiducials, at least 12 fiducials are recommended for robust coordinate training. Similarity transformations do not correct for shearing of images, so the field of view must remain normal to the sample surface during image acquisition and MS analysis.

Generally, a fiducial is located in the microscope image and the instrument system with the assistance of an integrated video camera. When the microprobe is positioned over the center of a fiducial mark, the same location is selected on the image in microMS with a right mouse click. This opens a popup window requesting the physical x, y position. A default x, y position is displayed in the popup which directly reads the stage position, pastes text from the computer clipboard, or predicts the closest location, depending on the selected instrument.

Fiducials are displayed on the microscope image as blue circles with labels corresponding to the nearest set position on the instrument, as shown in the schematic of Figure 3.4. A few feedback features are included to help the user assess the quality of the current registration. If applicable, labels display set points of the instrument coordinate system. The labels shown in

Figure 3.4B correspond to a Bruker MTP slide II adapter. For some instruments, a set of

71 preprogrammed positions are displayed, showing the predicted location of those points on the microscope image. A large deviation, typically due to an inaccurate input will be detected by a discrepancy in the expected and displayed label. Finally, the fiducial with the worst fiducial localization error is highlighted in red, indicating that specific position assignment should be reconsidered. Correcting the problematic fiducial will cause the next worst fiducial to be highlighted. Once the same fiducial stays highlighted, the registration is close to optimal.

Adjusting the worst fiducial is good practice to produce accurate targets.

With a full set of fiducials, the target positions may be saved in instrument-specific format for offline analysis. Alternatively, microMS can communicate directly with an instrument to instruct it to move and perform an analysis. Due to limited vendor support, direct instrument control is demonstrated on a lab-built stage.

The current microMS distribution supports the Bruker ultrafleXtreme, Bruker solariX,

AB Sciex oMALDI server, and a lab-built liquid microjunction extraction stage. For the supported MS systems, the user registers fiducials and saves instrument positions without modifications. Furthermore, microMS has ample room for customization due to the abstract base class construction of the CoordinateMappers package. The general framework remains unchanged, but there are opportunities to tune each function to a specific application. Examples include the ability to directly read fiducial positions from an instrument, grab the contents of the computer clipboard, and perform stage movement slop correction.

Accuracy of point-based similarity registration.

A vital metric for optically guided profiling is the target localization error. An accurate transformation between optical image and physical location ensures that each sample corresponds to the position of interest. microMS allows training sets of arbitrary sizes. Including

72 more fiducials reduces target localization error, effectively distributing uncertainty of a given fiducial over the entire transformation. Several factors influence accuracy including the precision of stage movement, fiducial localization accuracy (in both image and physical coordinate systems), number of fiducials, whole-slide image stitching, and proper sample positioning during image and MS acquisition. Users should carefully consider these factors to establish adequate probe size and distance cutoffs prior to data acquisition. To assess the accuracy of an MS system with microMS, an image based method was developed to link the requested and actual target positions, as shown in Figure 3.5. The target localization error is defined as the Euclidean distance between a requested position and the actual, transformed position during coordinate registration and is synonymous with accuracy. To assess this value, a thin layer of DHB matrix was coated on an ITO glass slide to act as a tracer for the probe position. A standard sample was prepared with 16-24 fiducials along the exterior of a target. An additional set of fiducials were included within this region to mimic the location of samples in a profiling experiment. The interior marks were not used for coordinate registration, but rather to assist with overlaying the pre- and post-analysis images. Several target locations were manually placed around each interior fiducial. Dividing the targets between multiple trials is useful for designing experiments to test the effect of possible confounding variables.

Next, fiducial training was performed with the MS system and the set of targets was desorbed with sufficient time to noticeably remove the DHB matrix. After desorption, the target area was optically imaged again to reveal the actual position of sampling events. Desorption locations and sizes are marked as blobs in microMS and saved for further analysis. To overlay the two images, subsets of the pre- and post-desorption image were cropped and roughly positioned prior to intensity-based registration with custom scripts in MATLAB (R2015b). The

73 resulting transformation was used to map the target pixel positions onto the post-extraction image. The distance between the requested and actual desorption positions is a direct measurement of the target localization error. With this method, the target localization of the

Bruker ultrafleXtreme was found to be 38.3 ± 3.9 µm (mean ± S.E.M, n = 71, Figure 3.5), while the liquid microjunction extraction stage accuracy is 42.8 ± 2.3 µm (n = 48; data not shown) over an area of approximately half a microscope slide, an error of about one part per thousand. As previously mentioned, the probe radius should be as large as the target localization error and the distance filter applied should be larger than the sum of this error and the probe radius.

In experiments to assess the effect of various confounding factors on target accuracy, desorption was repeated multiple times with the same sample and slide image. Different laser spot sizes, users, target locations and fiducial training sets were examined. The only significant factor found was the fiducial training set (Figure 3.5D). Overall accuracy is not dependent on the target location (Figure 3.5E), laser spot size or user (data not shown). Within an experiment, the accuracy is fairly constant, independent of the user or location on the sample. However, repeating an experiment with the same sample could produce significantly different accuracy.

This result confirms the profound effect of quality fiducial training sets on the target accuracy.

Extreme care is required when training fiducials to ensure the image target locations correspond to the expected mass spectra.

A demonstration: Sequential analysis of the same target.

For single cell profiling experiments, the physical location of a cell on the slide effectively isolates it from neighbors and prevents mixing, which greatly simplifies data fusion. microMS provides a utility for performing sequential analysis of the same target on different instruments with ease. Using the optical image as a map to record each target address, the image

74 position can be transformed into any supported instrument coordinate system. A careful selection of the order of experiments facilitates the repeated analysis of a sample to provide complementary chemical information.

Figure 6 shows two examples of sequential single cell profiling using MS instruments with different capabilities. In panel A, dispersed, rat cerebellum cells were initially profiled with a Bruker ultrafleXtreme to rapidly assess the lipid content with moderate resolution and mass accuracy. From the initial mass spectral dataset, cells without significant lipid signals are discarded from further consideration as they likely represent artifacts from optical imaging or sample preparation such as dust particles. The resulting population is then selected for follow-up, high resolution, high mass accuracy analysis with a Bruker solariX FT-ICR. Due to the increased sample acquisition time, exhaustive analysis of large populations is cost-prohibitive. Performing a preliminary filtering maximizes the efficiency of subsequent data analysis, without consuming the entire cellular content.

While the overall lipid profiles are similar, there are some discrepancies between lipid ratios of the two methods. This could represent changes in the sample layers that each technique is analyzing. Nonetheless, the advantage of single cell FT-ICR is immediately apparent with the ppm mass accuracy and over an order of magnitude higher mass resolution, shown in each inset for putative [PC(32:0)+H]+. Once the MALDI-TOF has identified cells with abundant lipid signal, they are filtered to locate individuals requiring exact mass measurement for elemental composition analysis. Such a workflow facilitates exhaustive cell population analysis while efficiently utilizing the FT-ICR as needed.

+ As a second example, Figure 3.6B displays a C60 -SIMS mass spectrum and the corresponding MALDI-TOF mass spectrum of a single cell derived from the rat suprachiasmatic

75 nucleus. Here, the low sample consumption of SIMS was leveraged by follow-up MALDI-TOF to provide more, complementary information than would be possible with either technique alone.

In the low mass range, peaks corresponding to phosphatidylcholine headgroup and a cholesterol fragment are apparent in the SIMS spectrum at m/z 184.09 and 369.31 respectively. MALDI-

TOF demonstrates better sensitivity to intact lipids and detects lipid dimers and peptides.

Comparing the identity of lipids over the same range, SIMS appears to favor sodiated adducts

([PC(32:0)+Na]+ and [PC(34:1)+Na]+ at m/z 756.47 and 782.55 respectively) more than the protonated forms seen in MALDI-TOF ([PC(32:0)+H]+ and [PC(34:1)+H]+ at m/z 734.54 and

760.55 respectively). These relative intensities likely reflect the different ionization processes occurring in each instrument. Together, a wide mass range is covered to provide a more complete profile of the sample.

These examples may represent the first demonstration of multiple MS platforms measuring the same individual cells with high throughput. In each example, the ability to repeatedly analyze the same cell was leveraged to acquire complementary information from multiple instruments. Such an experiment would be difficult to perform at high throughput without linking the target locations by the optical image of each sample. With microMS, sequential analysis is facile, enabling each cell to be exhaustively characterized by multiple techniques.

3.4 Conclusions and Future Directions

microMS is the first generation of an open source python package for robust image analysis and coordinate registration, which are essential for optically-guided MS profiling. microMS provides a rich feature set for image analysis suited for optically-guided MS profiling.

Targets may be automatically located, filtered, stratified and patterned prior to MS analysis.

76

These functions provide access to single cell profiling with multichannel fluorescence image analysis. The unique aspect of microMS is how mass spectrometers are represented for MS profiling. The implementation of specific MS systems through an abstract base class and software architecture provides a straightforward means for adapting microMS to arbitrary microprobe instruments. While this simplifies connecting microMS to new systems, it also facilitates sequential analysis of the same target by uniquely addressing each cell coordinate. We believe the rich feature set and ease of extending microMS to a variety of mass spectrometers and other instruments will facilitate the growth of single cell profiling.

3.5 Acknowledgments

We thank Stanislav Rubakhin for assistance with sample preparation, and Tong Si for useful discussions. We also gratefully acknowledge support from the National Institutes of

Health, Award Number P30 DA018310 from the National Institute on Drug Abuse and from the

National Institute of Mental Health Award Number 1U01 MH109062. TJC and EKN acknowledge support from the National Science Foundation Graduate Research Fellowship

Program and the Springborn Fellowship. TJC received additional support through the Training

Program at Chemistry-Interface with Biology (T32 GM070421).

77

3.6 Figures

Figure 3.1 Partial unified modeling language diagram of microMS class structure. Each experiment is contained in a microMSModel object, consisting of a list of targets (blobList), a microscopy image (slideWrapper), and an instrument mapper (coordinateMapper).

CoordianteMapper defines the set of instructions required for each inheriting class. Four concrete instrument implementations are provided with microMS.

78

Figure 3.2 Overview of blob finding with microMS. From an input image (A), the pixel intensity is filtered by a threshold intensity. B) Pixels above the threshold are grouped with neighboring pixels to generate putative blobs. Each group of pixels is evaluated for its size (in pixels) and circularity. The area (A) and perimeter (P) are directly measured from the threshold image.

79

Figure 3.3 Population-level filtering through the histogram window. A) A population of found blobs may be filtered by size, circularity, minimum pairwise distance, or fluorescence intensity.

The histogram can be divided into a low pass, high pass, or single interval with the appropriate blobs dynamically colored in the microscope image (B and C).

80

Figure 3.4 Schematic of fiducial training. A) The input image includes several fiducial points, such as etched x marks on the glass slide. B) An initial attempt at registration with labels of the nearest named coordinate for each fiducial. Fiducials are shown in blue, except the point with the worst fiducial localization error, which is in red. A set of predicted locations in yellow are also be displayed. C) After removing and retraining the worst fiducial the next worst fiducial is dynamically highlighted.

81

Figure 3.5 Determination of target localization error. A) Target locations (green) are marked around an image of an etched x mark. The sample is then coated with a thin layer of MALDI matrix and analyzed by optically-guided MS to generate desorption craters in the matrix. B) The location of resulting desorption events (red) are determined by optical microscopy. C) Image registration of panels A and B allows the direct mapping of requested target locations onto the desorption marks. Overlap is shown in yellow. The distance between these positions is the target localization error of the registration set. The effect of various parameters may be assessed simultaneously including multiple training sets, location on slide, or size of microprobe, as shown here. A three-way linear ANOVA demonstrated that while the specific fiducial training

82

Figure 3.5 (cont.) set significantly affected accuracy (p << 0.05), the location on the slide (p =

0.6) and spot size (p = 0.3) did not.

83

Figure 3.6 Sequential analysis of the same cell with two separate MS systems. Once a cell has been located in the optical image (top), its location remains fixed through multiple analyses allowing two instruments to probe the same set of selected cells. A) MALDI-TOF MS (middle) of a cerebellum-derived cell followed by MALDI-FT-ICR MS (bottom). MALDI-TOF provides high throughput screening of thousands of cells to highlight rare or representative individuals.

Here, FT-ICR provides high mass resolution and high mass accuracy for unequivocal elemental composition of selected cellular contents. B) SIMS profiling (middle) followed by MALDI-TOF

MS (bottom) with a DHB-coated, suprachiasmatic nucleus derived cell. SIMS provides

84

Figure 3.6 (cont.) information on small molecule compounds while MALDI-TOF MS effectively detects larger species, such as lipid dimers and peptides. The inset demonstrates some overlap of intact lipid coverage from each modality.

85

3.7 References

(1) Lanni, E. J.; Rubakhin, S. S.; Sweedler, J. V. J Proteomics 2012, 75, 5036.

(2) Rubakhin, S. S.; Lanni, E. J.; Sweedler, J. V. Current Opinion in Biotechnology 2013, 24, 95.

(3) Comi, T. J.; Do, T. D.; Rubakhin, S. S.; Sweedler, J. V. Journal of the American Chemical

Society 2017.

(4) Armbrecht, L.; Dittrich, P. S. Analytical Chemistry 2017, 89, 2.

(5) Chen, X.; Love, J. C.; Navin, N. E.; Pachter, L.; Stubbington, M. J. T.; Svensson, V.;

Sweedler, J. V.; Teichmann, S. A. Nat Biotech 2016, 34, 1111.

(6) Zenobi, R. Science 2013, 342, 1243259.

(7) Korte, A. R.; Yandeau-Nelson, M. D.; Nikolau, B. J.; Lee, Y. J. Anal Bioanal Chem 2015,

407, 2301.

(8) Kompauer, M.; Heiles, S.; Spengler, B. Nat Meth 2016, advance online publication.

(9) Lee, J. K.; Jansson, E. T.; Nam, H. G.; Zare, R. N. Analytical Chemistry 2016, 88, 5453.

(10) Passarelli, M. K.; Ewing, A. G. Current opinion in chemical biology 2013, 17, 854.

(11) Lanni, E. J.; Dunham, S. J.; Nemes, P.; Rubakhin, S. S.; Sweedler, J. V. Journal of The

American Society for Mass Spectrometry 2014, 25, 1897.

(12) Yeager, A. N.; Weber, P. K.; Kraft, M. L. Biointerphases 2016, 11, 02A309.

(13) Spitzer, Matthew H.; Nolan, Garry P. Cell 2016, 165, 780.

(14) Pabst, M.; Fagerer, S. R.; Kohling, R.; Kuster, S. K.; Steinhoff, R.; Badertscher, M.; Wahl,

F.; Dittrich, P. S.; Jefimovs, K.; Zenobi, R. Anal Chem 2013, 85, 9771.

(15) Ibáñez, A. J.; Fagerer, S. R.; Schmidt, A. M.; Urban, P. L.; Jefimovs, K.; Geiger, P.;

Dechant, R.; Heinemann, M.; Zenobi, R. Proceedings of the National Academy of

Sciences 2013, 110, 8790.

86

(16) Krismer, J.; Sobek, J.; Steinhoff, R. F.; Fagerer, S. R.; Pabst, M.; Zenobi, R. Applied and

environmental microbiology 2015, 81, 5546.

(17) Fagerer, S.; Schmid, T.; Ibanez, A.; Pabst, M.; Steinhoff, R.; Jefimovs, K.; Urban, P.;

Zenobi, R. Analyst 2013.

(18) Ong, T. H.; Kissick, D. J.; Jansson, E. T.; Comi, T. J.; Romanova, E. V.; Rubakhin, S. S.;

Sweedler, J. V. Anal Chem 2015.

(19) Jansson, E. T.; Comi, T. J.; Rubakhin, S. S.; Sweedler, J. V. ACS Chemical Biology 2016,

11, 2588.

(20) Pan, N.; Rao, W.; Liu, R.; Kothapalli, N.; Burgett, A.; Yang, Z. Planta Medica 2015, 81,

IL55.

(21) Pan, N.; Rao, W.; Kothapalli, N. R.; Liu, R.; Burgett, A. W.; Yang, Z. Analytical Chemistry

2014, 86, 9376.

(22) Laskin, J.; Heath, B. S.; Roach, P. J.; Cazares, L.; Semmes, O. J. Analytical Chemistry 2012,

84, 141.

(23) Li, B.; Comi, T. J.; Si, T.; Dunham, S. J.; Sweedler, J. V. Journal of Mass Spectrometry

2016, 51, 1030.

(24) Do, T. D.; Comi, T. J.; Dunham, S. J. B.; Rubakhin, S. S.; Sweedler, J. V. Analytical

Chemistry 2017.

87

CHAPTER 4

OPTICALLY GUIDED SINGLE CELL MASS SPECTROMETRY OF RAT DRG TO PROFILE

LIPIDS, PEPTIDES AND PROTEINS

Notes and Acknowledgements

This chapter is adapted from a completed manuscript coauthored by Thanh D. Do, Joseph

F. Ellis, Troy J. Comi, Emily G. Tillmaand, Ashley E. Lenhart, Stanislav S. Rubakhin, and

Jonathan V. Sweedler, submitted for publication on December 21, 2017 and accepted March 15,

2018 in ChemPhysChem (DOI: 10.1002/cphc.201701364). T.D. Do and J.F. Ellis wrote a majority of the manuscript with contributions from the other authors. T.D. Do performed

MALDI TOF MS and J.F. Ellis did the data analysis. T.D. Do, E.G. Tillmaand, A.E. Lenhart, and S.S. Rubakhin performed most sample preparation. T.J. Comi provided input on data analysis. E.G. Tillmaand and J.V. Sweedler provided the concept. J.V. Sweedler also provided funding, and aided in manuscript preparation. The authors gratefully acknowledge support from the National Institutes of Health, Award Number P30 DA018310 from the National Institute on

Drug Abuse, and from the National Institute of Mental Health, Award Number 1U01

MH109062, as well as the National Science Foundation Award No. CHE 16‐06791. E.K.

Neumann and T.J. Comi acknowledge support from the National Science Foundation Graduate

Research Fellowship Program. E.K. Neumann, J.F. Ellis and T.J. Comi acknowledge support from the Springborn Fellowship. T.J. Comi received additional support through the Training

Program at Chemistry-Interface with Biology (T32 GM070421).

4.1 Introduction

Dorsal root ganglia (DRG) consist of neuronal cell bodies, myelinated and non- myelinated axons (here termed neurites), cells related to vasculature and blood, immune cells, as

88 well as several types of glia, including satellite cells. DRG are vital for conveying sensory information from the periphery to the spinal cord and brain stem as part of the peripheral sensory-motor system in mammals.[1,2] Because DRG integrate and discriminate diverse types of sensory inputs, such as proprioception and nociception, efforts to identify and characterize the cell types dedicated to relaying specific kinds of sensory information are critical for understanding fundamental mechanisms of the sensory-motor system. In general, DRG cell populations can be stratified by their anatomy, morphology, and electrophysiology, and also via a range of intracellular markers, from patterns of gene expression to proteome, peptidome, lipidome, and metabolome heterogeneity. In 1978, Fields et al.[3] utilized cell surface markers to classify DRG cellular populations into four cell types: neurons, Schwann cells, fibroblasts, and unidentified flat cells. McMahon and Priestley[1,4] divided DRG cells into a minimum of three major subclasses based on size and neurochemical content, consisting of small cells

(predominantly nociceptors,[2] 10–30 µm) having unmyelinated axons that either (a) express calcitonin gene-related peptide (CGRP) or (b) bind to the lectin Griffonia simplicifolia IB4, and

(c) larger cells (25–60 µm in diameter) having high levels of neurofilaments and myelinated axons. The three populations overlap (as some cells belong to multiple groups), suggesting the presence of additional cellular subpopulations. Recently, multiple research groups have revealed more details about the molecular and functional diversity of DRG neurons with the application of single cell transcriptomics/RNA sequencing.[5-7] Notably, Usoskin et al.[7] performed single cell transcriptomics to identify 11 cell types in DRG, two of which were peptidergic. Our MS imaging study of DRG chemical profiles highlighted striking chemical complexity, including the presence of distinct morphologically defined spatio-chemical regions,[8] although at the level of

89 tissue sections. Here, we report label-free, multiplexed detection of endogenous lipids, peptides, and small proteins in populations of individual rat DRG cells.

Single cell metabolomics and peptidomics aim to recognize the similarities and distinctions between individual cells by linking their chemical dynamic profiles to cellular fate, function, homeostatic balance, and other biological phenomena.[9-12] Among the bioanalytical techniques used in single cell metabolomic and peptidomic investigations, mass spectrometry

(MS) is at the forefront, owing to its high analyte coverage, low limits of detection (LOD), versatile analyte sampling methods, and unique ability to be coupled with in situ or off-line orthogonal characterizations.[9, 11-20] MS imaging (MSI) focuses on direct analysis of tissue sections to determine the relative abundance and spatial distribution of analytes in tissue sections.[11, 21-25] In addition, multimodal MSI approaches have facilitated detection of diverse analyte classes from the same samples by utilizing different matrices and sequential tissue imaging.[26] Single cell chemical imaging remains relatively specialized due to the limited number of MS technologies capable of routinely providing micron resolution;[25, 27-33] discriminating a typical mammalian cell within an intact tissue slice requires the footprint of the

MS probe to be less than 3 µm.[32] To circumvent the challenges of cellular and sub-cellular imaging, several MS-based approaches have been developed. Low density populations of individual cells produced by enzymatic dissociation of different tissues can be deposited on a substrate and processed for MS analysis, significantly reducing cell-to-cell cross contamination, even when using microprobes exceeding cell sizes, albeit at the expense of native tissue context.

Populations of individual cells are suitable for subsequent high-throughput MS analysis enabled by fabricated microwell devices[34, 35] and optical imaging.[20, 36-39] Recently, we demonstrated a unique optically guided single cell MS approach to profile hundreds to thousands of cells from

90 different tissues and organs in a single experiment using both secondary ion MS (SIMS)[38] and matrix-assisted laser desorption / ionization (MALDI) MS.[37, 39, 40]

In the current work, we performed mass spectral classification of cell types from sequential MALDI MS analyses of the same cells, revealing heterogeneity in peptides and small proteins. The majority of the detected peptides were tentatively identified in a prior peptidomic study using liquid chromatography (LC) coupled with electrospray ionization (ESI) Fourier transform-ion cyclotron resonance (FT-ICR) MS and tandem MS (MS/MS), in combination with direct tissue analysis using MALDI-TOF MS.[41] Previously unreported signals that appear to be peptides were detected from a rare cell type.

4.2 Experimental Section

Chemicals and Matrix Preparation

All chemicals except where stated otherwise were purchased from MilliporeSigma (St.

Louis, MO) and used without further purification. DHB matrix solution was prepared by dissolving DHB (99% purity) to 50 mg/mL in 1:1 (v/v) LC-grade ethanol/water and 0.1% trifluoroacetic acid (TFA) solvent.

Sample Preparation

Single cell samples and tissue extracts were prepared from a total of three 2.5–3 month- old male Sprague-Dawley outbred rats (Rattus norvegicus) (www.envigo.com). The animals were housed on a 12-h light cycle and fed ad libitum. Animal euthanasia was performed in accordance with the appropriate institutional animal care guidelines (the Illinois Institutional

Animal Care and Use Committee), and in full compliance with federal guidelines for the humane care and treatment of animals. The studies were planned in accordance with the ARRIVE guidelines.[60] Rats were killed by quick decapitation using a sharp guillotine. Rat trunks were

91 placed on ice, where all surgical procedures were performed. Ten to twenty DRG per animal were surgically isolated during the ~10-min dissection procedure and placed into ~5 mL of cold modified Gey’s balanced salt solution (mGBSS) containing (in mM): 1.5 CaCl2, 4.9 KCl, 0.2

KH2PO4, 11 MgCl2, 0.3 MgSO4, 138 NaCl, 27.7 NaHCO3, 0.8 Na2HPO4, 25 HEPES, and 10 glucose, pH 7.2.

Single cell dissociation.

The DRG were incubated in 300 µL of 0.25% collagenase P in oxygenated Hank’s balance salt solution (HBSS) for 90 min to remove the surrounding connective tissue and isolate individual neurons. The sample was then centrifuged at 11,000 rpm for 3 min and the supernatant removed. The resulting pellet was washed with 500 µL HBSS. Next, the DRG pellet was treated with 500 µL of 0.25% trypsin in HBSS for 20 min, followed by centrifugation at

11,000 rpm to remove the supernatant. Finally, the DRG pellet was placed in 300 µL of 1% fetal bovine serum (FBS) in HBSS containing 33 µg/mL Hoechst 33342 nuclear stain (Thermo Fisher

Scientific, Waltham, MA; 10 µL of 1 mg/mL stock solution). Following a 20-min incubation,

DRG were mechanically dissociated by trituration. Cells were stabilized using an equal volume

(300 µL) of a mixture of 80% glycerol, 20% mGBSS (v/v), and after 5–10 min, plated on ITO- coated glass slides (25 mm × 75 mm). The samples were then stored at room temperature, in the dark overnight, to allow the cells to adhere onto the ITO-glass surface. Excess glycerol- containing media was removed from the preparations shortly before analysis and the entire substrate was washed with 1 mL of 150 mM ammonium acetate (pH = 10). This wash removes excess glycerol and does not induce observable damage to the cells.[37-40] Optical images of the substrate and cell populations were obtained, followed by matrix application and MS analysis.

92

All steps following washing were performed on the same day to prevent sample degradation and are detailed below.

Samples from each animal were treated in separate Eppendorf tubes (1.5 mL) and divided across a total of four ITO-slides; each slide contained cells from every animal in discrete locations to assess possible batch effects.

Tissue extraction.

Isolated DRG were transferred into 1 mL of aqueous salt solution containing 0.15 M

o NaCl, 1 mM CaCl2, 1 mM HEPES for 4 h at 4 C. The sample was then centrifuged at 6000 rpm for 1 h and the supernatant collected. The supernatant was salted out with dry (NH4)2SO4 in the presence of 0.01 M EDTA and stored for three days at 4oC. The obtained solution was then centrifuged at 11,000 rpm for 1 h and the supernatant was separated from the pellet. The supernatant was desalted by solid phase extraction using a Discovery DSC-18 52603-U tube

(MilliporeSigma). The resulting samples were aliquoted and freeze dried using a miVac Duo-

Concentrator (GeneVac, Ipswich, UK]).

Optical Imaging and Determination of Sample Coordinates for Individual Cells

Each dispersed cell population on an ITO-coated glass slide was imaged using an Axio

Imager M2 (Carl Zeiss, Oberkochen, Germany) in fluorescence and brightfield modes. An X-

CITE 120 mercury lamp (Lumen Dynamics, Mississauga, Canada) and a 31000v2 DAPI filter set (Chroma Technology, Irvine, CA) were employed for fluorescence imaging. A 10× objective was used to obtain a mosaic image of the targeted surface with 13% overlap between neighboring images. Images were taken using an AxioCam ICC camera (Carl Zeiss) with a resolution of 1936 × 1460 pixels. The final image was loaded into the lab-built microMS software[56] (available at (http://neuroproteomics.scs.illinois.edu/microMS.htm) for either manual

93 or automatic cell finding. Individual cells (50–200 cells per animal in each replicate) were manually selected from whole slide microscope images. Each cell was selected based on both fluorescence and brightfield images. In some cases, if the fluorescent signal was faint, the cells were still selected if the brightfield image showed a well-defined cell shape. Fiducial marks were used to register the image coordinates to the x,y translation stage coordinate of the mass spectrometers via the microMS software.[56] On the basis of the registration, the cell coordinates were saved in a custom geometry file for the MALDI MS FlexControl software (Bruker,

Billerica, MA) for mass spectral acquisition.

Matrix Application

ITO-coated glass slides with dispersed cells were affixed onto a rotating plate for automatic matrix application, as described elsewhere.[40] The distance between the spray tip and the rotating plate was 2 cm, with a nitrogen gas pressure of 50 psi. The solution flow rate was set to10 mL/h, resulting in a matrix coating of 15 mg/cm2, respectively.

MS Measurements of Substance P Standards using CHCA and DHB

Substance P standards were prepared by serial dilution of the stock solution (1 mg/mL in water) to the desired concentrations (100 pM to 1 fM). The matrices for triple-layer application were prepared followed a previous protocol by Li and co-workers.[61,62] The first layer was formed by pipetting 0.5 µL of 100 mM CHCA in 40/60 v/v methanol/acetone on a polished

MALDI steel plate, and allowed to crystallize. The second layer was formed by adding 0.4 µL of saturated CHCA in 30/70 v/v methanol/water. After the second layer was crystallized, 1 µL of substance P standard was plated in the center of the matrix spot and allowed to dry. A gentle wash was then performed using 0.1% TFA in water for 15 s; the liquid was blown off with compressed air. For the DHB dried droplet approach, 1 µL of DHB matrix was plated on the

94 same substrate, immediately followed by 1 µL of substance P standard. The sample spot was allowed to crystallize before analysis.

MS Instrumentation

An ultrafleXtreme MALDI TOF/TOF mass spectrometer (Bruker) with a frequency tripled Nd:YAG solid state laser was used for cell analyte profiling using the optically guided approach. Single cell MALDI MS analysis was performed in two stages. In the first stage (LM analysis), the molecular mass scan window was set to m/z 400−6000 and the laser was operated in the “Ultra” mode, producing a ∼100-μm diameter footprint. Each spectrum obtained in TOF reflectron mode represents the summed signals acquired during 1,000 laser shots fired at 1 kHz.

In the second stage (HM analysis), the molecular mass window was working in linear TOF mode. The mass range was set to m/z 4,000–20,000 and the laser was operated in the “Small” mode (~50 µm) footprint. The FlexControl AutoXecute feature was utilized with the custom geometry file as previously reported.[37-40]

A 7T SolariX XR FT-ICR mass spectrometer (Bruker) equipped with a dual ESI/MALDI source and managed by the ftmsControl software (v. 2.1.0, Bruker) was used to measure the molecular masses of peptides in the tissue extracts at higher resolution. External linear calibration was performed using perfluoroheptanoic acid (PFHA) clusters prior to MALDI MS analysis. Spectra were acquired in a mass range from 500–15,000, yielding a transient length of

2.4 s. Instrumental parameters of importance include a source funnel RF amplitude of 300 Vpp,

TOF delay of 2.8 ms, accumulation hexapole of 1.4 MHz and 2000 Vpp, and ICR sweep excitation of 20%. Spectra were acquired with 150 laser shots using the small laser setting (~40-

µm footprint) at 50–80% laser power and a frequency of 1000 Hz. Data was recalibrated in

95

DataAnalysis (v. 4.4, Build 200.55.2969, Bruker) using internal quadratic calibration with PFHA clusters for accurate mass values.

Data Analysis

Acquired spectra were imported into MATLAB (MathWorks, Inc., Natick, MA). Spectral preprocessing consisted of baseline correction, noise-smoothing, normalization, and resampling to a bin width of 0.5 m/z. Mass spectra were normalized with the msnorm function in MATLAB to standardize the area under the curve to the group median. PCA was performed on all cells

(both LM and HM) for m/z 400–6,000 (Figure 4.3) and 4,000–20,000, respectively. Peaks of highest variance for each range were selected for targeted spectral-type comparisons. Cells were considered to contain the m/z value if they exhibited a spectral peak with a S/N >3, otherwise the intensity for that m/z was set to 0 for successive analyses. The correlation heatmap was generated with all pairwise combinations of each peptide marker for all cells. The correlations, , were plotted as blue and red gradients for positive and negative correlations, respectively ( ≤ 0.05).

Correlations not significantly different than 0 ( > 0.05) are represented as white, whereas the intensity of blue and red represents the magnitude of correlation between peptide markers. The correlation values were clustered by HCA using an unweighted average distance to produce groups of m/z values. Peptide markers possessing both a statistically significant positive correlation within the population and negative correlation ( ≤ 0.05) to those outside of the population were selected as representative peptide markers for A-type, B-type, and C-type cells.

96

4.3 Results and Discussion

Optically Guided Single Cell MALDI MS of DRG Cell Populations for High-Throughput

Profiling of Lipids and Peptides

Single cells from enzymatically dissociated DRG from three animals were prepared following an established protocol.[37-40] The dissociated cells were placed onto four indium tin oxide (ITO)-coated conductive glass substrates in three separated regions marked for samples from individual animals (see Figure 4.1). Additional details can be found in the Experimental section. Each cell was sequentially interrogated twice with a Bruker ultrafleXtreme MALDI-

TOF/TOF mass spectrometer. The first set of measurements utilized a 100-µm laser spot size and

MS acquisition in reflectron mode optimized for the m/z 400–6,000 mass range, referred to as low molecular (LM) mass analysis. These settings were used previously for single cell profiling of rat pituitary[37] and islets of Langerhans.[39, 40] Next, the same cells were individually re- analyzed with a 50-µm laser spot size with MS acquisition in linear detection mode, optimized for higher molecular weight molecules (m/z 4,000–20,000), and referred to as high molecular

(HM) mass analysis. The smaller beam size used for the HM analysis provides a higher laser fluence (at the same laser power and setting), which is necessary for desorbing and ionizing cellular analytes. To ensure that the same target spots/cells were assayed by both LM and HM analyses, the sample slide was not removed until both analyses were finished.

Figure 4.2 shows representative mass spectra of single DRG cells randomly selected from the LM analysis, all of which contain peaks corresponding to lipid monomers, dimers, and occasionally, peaks of lipid trimers and some peptides. The average lipid profile of DRG cells obtained from MALDI MS is shown in Figure 4.2E, which includes predominantly phosphatidylcholines (PCs). The same class of lipids was also primarily observed in a single cell

97

SIMS profiling study using an ionic liquid matrix.[38] We acknowledge that several matrices and nanoparticles have been designed to improve the detection of low-abundance lipid classes.[42]

However, since our goal was to detect lipids, peptides, and small proteins in the same cells, we used 2,5-dihydroxybenzoic acid (DHB) as the matrix. The signals with the most abundant intensities correspond to PC(34:1) at m/z 760.6, followed by PC(32:0) at m/z 734.6, and PC(36:1) at m/z 788.6. Similarly, lipid dimer peaks of homo- and heterodimers of PC(34:1) are among the most intense for the m/z 1200–1600 region, identified by accurate molecular masses and MS/MS data (see Supporting Information Figures S1 and S2). The MS/MS data were obtained by randomly ablating a sample surface region containing many cells or cell clusters. Of note, the signal intensity of the heterodimer of PC(36:1) with PC(34:1) is more abundant than that of the heterodimer of PC(32:0) with PC(34:1), although their monomer signal intensities have the opposite intensity relation. Many cells also show mass spectral peaks above m/z 2,000, as discussed below.

MALDI MS Single Cell Profiling and FT-ICR MS on DRG Analyte Extracts Detects Peptides that Reflect DRG Cell Heterogeneity

The single cell mass spectra obtained from our LM analysis were subjected to principal component analysis (PCA), as shown in Figure 4.3. The score plot (Figure 4.3A) shows overlap of data points acquired from cells isolated from different animals, demonstrating low batch-to- batch variation. PC1 primarily separated cells based on differences in the intensity of lipid signals (68.8% of variation explained), partially reflecting methodological variation in cell targeting accuracy and cell morphology. The PC2 loading plot (Figure 4.3B) is populated with a series of mass spectral peaks tentatively assigned to peptides, but not lipids, based on their mass

98 range and isotopic distribution. We note that heterogeneity in matrix application and the overall weak intensities of these peaks lowers the precision of m/z measurements.

As such, some of the m/z values derived from the average profile (Figure 4.3B), especially those at high masses, correspond to the average molecular masses rather than monoisotopic masses. For example, the mass spectral peak at m/z 4406 in Figure 4.3B has a monoisotopic mass of 4404. However, there is no signal with m/z 4404 detected in our prior LC-

MS peptidomics work on tissue extracts and releasates.[41] Il’ina et al.[43] identified a group of peptides in Wistar rat brain extract, referred to as membrane-tropic homeostatic tissue-specific bioregulators, including a peptide at m/z 4403; unfortunately, the primary sequence of the peptide was not reported, and so this remains unknown. High-mass resolution MALDI MS measurements of the extracts were performed using a Bruker SolariX FT-ICR mass spectrometer. Our results are summarized in Table 1, where the observed molecular masses are matched to the list of peptides from our previous report.[41]

The FT-ICR MS results indicate that the isotopically resolved ion at m/z 4404.672 corresponds to the averaged peak at m/z 4406 in the TOF MS measurements. The intensity of m/z

4404.672 from the DRG extract was weak, preventing direct MS/MS sequencing. Accurate FT-

ICR MS data from this work, along with published information,[41] was used to support the identification of several other peptides observed in our single cell profiling experiments, including vimentin, thymosin beta isoforms, and neurofilaments, in which the peptides from the two latter classes of proteins were also detected in single cell measurements. Peptides derived from gamma-synuclein were also detected in the extracts, but were absent from the single cell data. Other mass spectral peaks detected in the single cell experiments were mass matched to the list of peptides detected in the prior LC-MS study[41]. Many of the peaks observed here are

99 consistent with signals of peptides derived from myelin protein P0 (m/z 1947, 2263, and 3989) and myelin basic protein S (m/z 4337). Myelin basic protein, the second most abundant protein in the central nervous system, is encoded in a large complex of oligodendrocyte lineage genes.[44]

Previous studies have shown that myelin proteins are found predominantly in myelinating glial and immune cells. Several functions are attributed to myelin proteins, including maintaining the adhesion of the cytosolic surface of multilayered compact myelin, and interacting with polyanionic proteins such as actin, tubulin, Ca2+-calmodulin, etc.[44, 45] Adult Schwann cells, on the other hand, express negligible levels of myelin glycoprotein P0.[46] Vimentin is another protein that has been found to be abundant in satellite glial cells,[47-49] and in neurons at the early stage of development and following injury.[50-52] Other peptides (m/z 2141, 2920, 3015, and

3093; see Table S1) are likely derived from neurofilament polypeptides found in DRG neurons.[53] Neurofilaments are involved in axonogenesis and neurite extension.[51, 52, 54]

Approximately 40% of the rat lumbar DRG neurons are characterized by a high content of phosphorylated heavy-chain neurofilaments.[1, 53, 54]

Dividing the cell population based on the PC2 score allows coarse classification of the cellular population. The mass spectra of cells found on the negative side of PC2 primarily contained peptide peaks in the m/z 2,000–3,000 region, without peaks at higher masses. These species, tentatively assigned to myelin and neurofilament proteins, were often detected together with signals from lipid trimers. One notable peak is m/z 2263, which falls within the lipid trimer mass range (~m/z 2,200–2,500). MS/MS was performed on m/z 2263 to deconvolve the possible contributions of lipid trimers having similar masses (±1 Da). The LIFT-TOF/TOF product-ion mass spectrum of m/z 2263 (±5 Da precursor mass selection window) obtained by random continuous acquisition of data from many cells and cell clusters yielded intense peaks

100 corresponding to lipids. The fragments are consistent with the precursor ions being a sodiated lipid trimer composed of a PC(34:1) dimer and another lipid with m/z 723.6 (neutral loss to produce a PC(34:1) dimer), for a theoretical m/z of 2264.72. Since it is not possible to isolate m/z

2263.27 (tentatively, a myelin protein P0 fragment) from m/z 2264.72 (a lipid) with the

TOF/TOF instrument used here, we cannot assign identity of m/z 2263 signal solely to a peptide fragment. However, the presence of several cells with intense peaks at m/z 2263 in the absence of any other lipid trimers suggests that another component is detectible and is unlikely to be a lipid trimer.

Cells with positive PC2 scores predominantly contained higher mass peaks assigned to peptides derived from thymosin beta-4 (m/z 4716) and myelin basic protein S (m/z 4337).

Previous immunohistochemical work showed that thymosin beta-4 is located in both the cell bodies and axons of neurons.[55] As mentioned above, peaks at m/z 4404 and 5487 did not match any entry in the peptide database generated by our prior LC-MS analysis of tissue extracts.[41]

Overall, the PC2 loading plot and direct survey of individual mass spectra indicate the presence of heterogeneity in the cellular peptide content, including molecules related to myelin proteins

(P0 and basic protein S), neurofilaments, thymosin beta-4, and the uncharacterized peptides at m/z 4404 and m/z 5487.

An alternative explanation for the observed heterogeneity is that the differences in peptide content are due to experimental artifacts; i.e., the sensitivity of detection of high mass peptides is disproportionately affected by spatial cell targeting accuracy, MALDI matrix deposition and crystallization, cell size and/or instrument parameters/performance. To investigate this possibility, we examined individual DRG mass spectra and microscopy images to determine how the MS data correlate with morphological differences (Figure 4.4). The four

101 examined cells were from the same sample slide, and assayed in the same batch using identical instrument settings. In this case, similar targeting accuracy was achieved.[39, 56] Despite the similarity of the methodological parameters, all four cells displayed mass spectral heterogeneity.

Furthermore, cells with mass spectra containing the same major peaks as those shown in Figure

4.4 were reproducibly found in data sets across animals and batches. These spectra indicate that the observed qualitative differences reflect biological phenotypes rather than analytical uncertainties. Focusing on the representative cells, cell A (Figure 4.4A and 4.4E-row 1) is non- circular with a diameter of 15 µm. Its mass spectrum shows intense lipid signals without any peaks above m/z 2,000, except m/z 2264 (myelin protein P0 or a lipid trimer). The size and mostly undetectable peptide content of cell A makes it a candidate for identification as a supporting cell, such as a glia. Cell B (Figure 4.4B and 4.4E-row 2) appears to have a nucleus that comprises most of its volume. Interestingly, the mass spectrum from cell B displays intense lipid signals, along with at least four major peaks at m/z 2031, 3777, 3989, and 4963. As discussed above, the peaks are tentatively assigned to vimentin, transitional endoplasmic reticulum ATPase (TER ATPase), myelin protein P0, and thymosin beta-4, respectively. Cells C and D display larger overall sizes with more distinct membrane boundaries, as seen in their brightfield images (Figure 4.4E, rows 3 and 4). The morphology of cell C is non-circular with two localized sites exhibiting intense Hoechst fluorescence. These images suggest either that cell

C maintained a connection with a supporting cell (marked with a red circle), or that it was mechanically damaged during deposition and part of the nucleus detached from its cell body, evidenced by the faint nuclear stain in the cell body. Cell D is also large and presents diffuse nuclear fluorescence, which we observed regularly with other types of large cells. Both of the latter cells produced mass spectra containing the same major peaks at m/z 4337 and 4404 (Figure

102

4.4C, D), suggesting that these tentative peptides are primarily found in large DRG cells.

Furthermore, due to the presence of neurofilament-related signals, it is likely that cells C and D are DRG neurons. In addition, cells C and D are 9.7 mm apart on the slide, eliminating the possibility that the detection of peptide signals, especially m/z 4337 and 4404, is biased by cell location on the slide.

Cells mechanically damaged during sample preparation should have limited effect on the analysis of cellular peptide content. Our work and that of others have demonstrated that treatment with MALDI matrix yields similar results to sample fixation.[57, 58] Cells that rupture would lose much of their peptide content; hence these cells would lack peptide peaks and be excluded from further analysis. Additionally, restricting analysis to cells sufficiently separated from each other ensures that cell-to-cell contamination does not occur. Accordingly, we conclude that the diverse morphology observed between cells is not an indication of disturbance caused by sample treatment.

Sequential MALDI MS Profiling Improves Analyte Coverage and Detection of High Molecular

Mass Species and Reveals Peptide Heterogeneity in Morphologically Similar Cells

As expected, cell morphology does not predict peptide content; cells with morphologies similar to cell D (Figure 4.4) did not necessarily contain a detectable amount of peptide at m/z

4404. Figure 4.5A–F presents mass spectra of three cells (cells E, F and G) with similar morphologies (roughly circular with a diameter of ~25 µm) to cell D (Figure 4.4). Cells E and F have a similar LM peptide content as cells C and D in Figure 44., with the pair of mass spectral peaks at m/z 4337 and m/z 4404, but very minor peaks within the m/z 2,000–3,000 range. On the other hand, cell G, while having a similar morphology, exhibits a chemical profile populated with intense signals of m/z 2142, 2264, and 3094, but no signals at higher masses. This cell’s

103 peptide content is different from the four cells shown in Figure 4, except for m/z 2264, which is common with cell A. Since m/z 2142 and 3094 are tentatively assigned to peptides from neurofilament heavy and medium proteins, respectively, cells G and H in Figure 4.5 represent a population of neurofilament-rich cells. As mentioned above, neurofilaments play a major role in the developmental regulation of neuronal types in rat DRG.[59]

We note that a major drawback of LM analysis lies in the low intensities of the mass spectral peaks within the m/z 4,000–6,000 mass range. To circumvent this issue, we performed

HM analysis with a method optimized for the m/z 4,000–20,000 mass range at the expense of mass resolution. Following LM analysis, the same cells were subjected to HM analysis to cover the analyte range from m/z 4,000–20,000. The relative signal intensities of peptide peaks in the overlapping region between LM and HM analysis (i.e., m/z 4,000–6,000) increased by at least an order of magnitude. For example, m/z 5487 is barely observed in cells C and D (Figure 4.4) and is apparently absent in cells E and F (Figure 4.5) after LM analysis. When the same cells were examined with HM analysis, the mass spectra showed intense signals of m/z 5487 and large signal enhancements for m/z 4337 and 4404. As shown in the right panels of Figure 4.5, the HM analysis reveals additional heterogeneity between cells E, F and cells G, H, including several smaller peptide signals in Figure 4.5B, D and polypeptides/small proteins in Figure 4.5F, H. The apparent peptide and protein content heterogeneity may reflect the existence of multiple DRG cell types, even for those with similar morphologies. Overall, these observations highlight the utility of sequential analyses on the same cells and specifically suggest that the DRG population may be stratified by peptides found exclusively in specific cellular groups. Nevertheless, a thorough, quantitative analysis is necessary to identify these peptide markers, as discussed below.

104

Classification of DRG Cells Based on Direct MALDI MS Peptide Profiling

The spectra of the representative cells shown above suggest a cellular classification scheme according to the intensities of selected mass spectral peaks. Visual examination of the mass spectral features shown in Figure 4.5 suggests that based on peptide content there are at least two DRG subpopulations present. Based on the PC2 loading plot (Figure 4.3), m/z values with high loadings were selected as potential peptide markers as they represent the highest source of variance in the dataset. The pairwise linear correlation coefficients, , were calculated between each putative biomarker for all cells, with < 0.05 indicating statistical significance.

The correlation matrix was plotted where blue and red represent positive and negative correlations, respectively, that are statistically significant. An ideal set of biomarkers would show high, positive correlations with all members of the set and large, negative correlations with all ions outside the set. The potential peptide markers were sorted with hierarchical clustering analysis (HCA), revealing roughly three populations of cells termed A-, B-, and C-type, defined by 19 markers (Figure 4.6A).

The peptide markers for type A include m/z 4337, 4404, 5487, and 4926 in both LM and

HM, and m/z 6295 in HM. For type B, the peptide markers are m/z 1947, 2031, 2142, 2263,

2792, and 3015 in LM, and m/z 7985 and 8269 in HM. For type C, the peptide markers are all in

HM: m/z 6666 and 9437, and the tentative proteins at 11.3, 12.5, 14 and 14.5 kDa. Some ions chosen by PCA were excluded due to their positive correlation with markers from other cell types, including m/z 3777 and 3989, which were excluded from type B due to positive correlation with markers in type C, and m/z 4078 and 4717 in type C as they show some significant overlap with markers in type A (Figure 4.6A).

105

With the set of ions selected, each cell was tested for membership in the respective group.

A cell was considered a member of a given type if the cell contained any peptide markers with a signal-to-noise ratio (S/N) >3. Overall, the peptide markers from three groups successfully classified 919 cells (78% of the total cells analyzed from three animals) into three populations:

498 cells in type A, 386 cells in type B, and 211 cells in type C. Additionally, 129 cells contained both markers of types A and B, 69 cells contained both markers of types A and C, and

67 cells contained both markers of types B and C (Figure 4.6B). Finally, 24 cells contained peptide markers of all three cell types. With this classification approach, cells sorted into a group may express a single marker, or multiple markers from both LM and HM analyses.

Additional insights are provided by comparing the cellular classifications utilizing markers of LM and HM analyses separately; 89% of the LM-A cells (133 out of 149 cells stratified by LM-A markers) exhibit the markers characteristic for HM-A cells. Since the type A classification is based on the same ions appearing in both LM and HM (e.g., m/z 4404), the significant overlap between LM-A and HM-A is consistent with positive correlation among the markers. The small disparity highlights the difference in optimization between the methods.

Other reasons for the disparity may include matrix depletion after LM analysis, leading to inadequate HM signal from some cells, or a decrease in ion suppression (due to lipid removal) after LM analysis, improving analyte detection in HM analysis of other cells. These reasons hold true for all cell types, not just type A. Nonetheless, most cells in LM-A are also in HM-A and vice versa, which indicates a fairly robust repeatability of measurements between these analyses.

Since the two analyses were performed sequentially, without removing the samples from the mass spectrometer, this difference is not due to spatial target accuracy.

106

The type B classification shows a larger disparity between HM and LM analyses, where

LM and HM markers do not share common peptides. Only 41% of cells (130 out of 315) in LM-

B are also in HM-B. The leaner overlap between the two classes indicates that type B is more heterogeneous.

The type C classification was missed in the original assessment using only mass spectral types (Figure 4.5). As mentioned above, the two biomarkers of LM-C, m/z 4078 and 4717, have a poor S/N in LM and positively correlate with markers of A-type (Figure 4.6A). The HM biomarkers, especially m/z 9437, are more prominent and HM analysis also shows enhanced signals of m/z 4078 and 4717. As a result, LM-C markers were unable to classify cells as C-type, whereas HM-C markers helped categorize 211 cells.

Taken together, the three cell types account for 78% of the cells analyzed (919 of 1171 cells from three animals). Cells containing none of the peptide markers are typified by a lack of any peptide signal. Presumably the peptide content of the unclassified cells is below the LOD of our instrument. Figure 4.6C shows a 3D scatter plot of the signal intensities of three distinct peptide markers (m/z 4404 for A-type, m/z 1947 for B-type, and m/z 9437 for C-type), showing good separation between classes for these m/z values.

The peptides or small proteins detected in the HM analysis were not characterized due to the limitations of the TOF/TOF MS/MS instrument at that mass range. Our prior LC-MS study[41] focused on smaller peptides, so sequencing would require additional peptidomic measurements optimized for higher molecular mass species. As a discovery research outcome, the mass spectral types are largely distinct and HCA on the marker correlation matrix demonstrates one method for identifying cellular biomarkers using just single cell MS data sets.

107

The data shown in Figures 4.5 and 4.6 also demonstrate that although MALDI MS consumes analytes, our results show that the remaining single cell material, including the associated amount of MALDI matrix, is sufficient to perform sequential analysis of the same cells. In previous studies on rat cerebellar cells, we demonstrated that MALDI MS measurements can be performed to gain additional chemical information after the cells are analyzed with

SIMS.[56] We have also off-line hyphenated MALDI MS with CE-MS via a liquid microjunction surface sampling probe to measure amino acids and dopamine in peptide-profiled individual rodent pancreatic islet α and β cells.[39] The present study illustrates the first example of sequential MALDI MS analysis for improved analyte coverage from lipids to peptides and small proteins at the single cell level.

Several peptides previously reported to be localized within the DRG were not detected, including CGRP, which is expressed in over half of small DRG neurons, along with neuropeptides such as substance P, somatostatin, vasoactive intestinal polypeptide, and galanin.[1,

4] Their absence in our measurements could be caused by analyte stability issues (proteolysis), ion suppression during analyte ionization and detection, analyte loss due to sample rinse, or analytes being present at levels below our LODs. Another possibility is that the fully processed peptides are mostly localized to cell processes that were not targeted in our experiments as we analyzed somata. Moreover, the choice of MALDI matrices may play a key role. For example, we note that substance P is not detected using the DHB dried droplet approach over a range of concentrations from 100 pM to 1 fM. However, when using triple-layer α-cyano-4- hydroxycinnamic acid (CHCA) matrix/sample preparation, we detect substance P over this concentration range (see Supporting Information Figure S12). Nevertheless, variation in matrix layers remains challenging for applications on biological samples, and thus, requires further

108 optimization. Regardless, the detected compounds allow cell type classification. Our work extends the analyte coverage for chemical profiling into the range of small proteins at the level of individual mammalian cells.

4.4 Summary and Conclusions

The heterogeneity and chemistry of individual cells are often hidden in the ensemble- average data sets resulting from traditional bulk analysis. Here, optically guided single cell MS profiling was utilized to investigate the chemical content of isolated DRG cells in a high- throughput manner. A unique property of the approach is sequential MALDI MS analyses; we assayed more than 1,000 individual DRG cells and identified lipids, peptides, and small proteins.

The chemical content of each cell is represented by two mass spectra, which together cover the mass range from m/z 400 to 20,000. The first LM analysis yielded mass spectral information about lipids and peptides up to m/z 6,000 with high resolution, and the second HM analysis captured high molecular weight polypeptides and small proteins in the same cell up to m/z

20,000. The mass spectra reflect the nature of cell types and morphologies. We also show that although MALDI MS is generally considered destructive for small samples, the remaining cellular material after the initial analysis was sufficient for a second follow-up MS analysis.

As applied to the rodent DRG, our workflow enabled detection of a number of analytes, including several peptides derived from myelin protein P0, myelin basic protein, neurofilaments, vimentin, and thymosin beta, at single cell levels. Small cells tend to contain fewer peptides than large cells, with myelin protein P0 and vimentin among the most common. These are proteins known to be expressed at high concentrations in glial cells.[44-47] In addition, cells appearing as structurally damaged with intact nuclei showed peptide signals putatively assigned to TER

ATPase and thymosin beta-4. We also detected uncharacterized putative peptides at m/z 4404

109 and m/z 5487, together with myelin basic protein S and neurofilament peptides in a subpopulation of large cells. Of note, peptides from neurofilament proteins, which contribute to axonogenesis and neurite growth,[50, 53, 54] were detected in many cells. The identities of these peptides were evaluated by measuring the exact molecular masses of peptides in a tissue extract via FT-ICR MS and matching them to corresponding theoretical data as well as previously published information on DRG neuron chemical profiles. These measurements also revealed more peptides derived from the aforementioned proteins.

The mass spectra were classified into three types, A, B and C, based on different peptide signal profiles, in which types B and C are more chemically heterogeneous than type A. This result could be due to the fact that type-A cells were selected based on the presence of the same ion (m/z 4404) in the results of both LM and HM analyses. We also detected small proteins at 14 kDa that appear to be specific for cell type B. These three mass spectral types account for 78% of the cells analyzed in samples from three animals. In a previous report, Usoskin et al.[7] classified

DRG based on neuropeptides, and showed that 11% of DRG cells are neuropeptidergic and express high levels of substance P (Tac1 gene), trkA (Ntrk1), and CGRP. In contrast, our approach is untargeted. Future work will aim to characterize neuropeptides at the single cell level and further refine our subpopulation classification.

Single cell techniques are often evaluated in terms of the number of cells that can be analyzed per experiment (the breadth of cellular profiling) and the number of detected analytes per cell (the depth of cellular profiling). We show that our optically guided MS technique not only assays a large number of cells, but also provides coverage of lipids, peptides, and small proteins in individual cells.

110

4.5 Acknowledgements

We gratefully acknowledge support from the National Institutes of Health, Award

Number P30 DA018310 from the National Institute on Drug Abuse, Award Number 1U01

MH109062 from the National Institute of Mental Health, and the National Science Foundation,

Award No. CHE 16-06791. We also acknowledge support from the National Science Foundation

Graduate Research Fellowship Program (T.J.C. and E.K.N.) and the Springborn Fellowship

(T.J.C., E.K.N. and J.F.E.). T.J.C. received additional support through the Training Program at

Chemistry-Interface with Biology (Grant T32 GM070421).

111

4.6 Figures and Table

Figure 4.1 Optically guided high-throughput single cell MS profiling of rat DRG cells. Intact

DRG were enzymatically dissociated and stained with a fluorescent nuclear dye. Isolated cells on

ITO-coated glass substrates are located with brightfield and fluorescence imaging and targeted for MS analysis.

112

Figure 4.2 A–D) Representative mass spectra in the LM range (m/z 400–6000) acquired from single DRG cells. Each inset highlights the m/z range above 2,000. E) Average lipid profile of

DRG cells; a) lipid monomers and b) lipid dimers.

113

Figure 4.3 PCA of MS data acquired from populations of single DRG cells. A) Score plot of data obtained from the LM analysis with 95% confidence ellipses. B) PC2 loading plot of the m/z

>2,000 region showing mass spectral peaks that vary the most in the analyzed data set according to PC2, including numerous peptides. “n” indicates the number of cells examined in each animal.

114

Figure 4.4 Examples of the biochemical and morphological variability of DRG cells. A–D)

Mass spectra of four different DRG cells; E) the corresponding cell fluorescence and brightfield images collected from the same animal. The scale bar for all images is 50 µm. The green circle approximates the LM analysis laser footprint (~100 µm). The red circle in the cell C image annotates the nucleus-like structure, as discussed in the main text. Mass spectral peaks are labeled with m/z values, where the asterisk indicates the average molecular mass.

115

Figure 4.5 Examples of the biochemical and morphological variability of DRG cells analyzed with sequential (LM followed by HM) MALDI MS profiling. A–J) Representative mass spectra obtained from sequential MALDI MS profiling of DRG cells. The combined fluorescence and brightfield images of the cells are shown in the right panels. The scale bar is 50 µm. Two

MALDI MS (low mass and high mass) analyses were performed on the same cell, corresponding mass spectra shown in the same row and color. Major mass spectral peaks are annotated with m/z values. The last cell (shown in orange, panels I, J), which is cell B from Figure 5.4, is morphologically and biochemically distinct from the other cells.

116

Figure 4.6 Classification of DRG cells in MALDI MS data sets. A) Heatmap represents covariance of intensities for m/z values selected from PC2 and sorted via HCA. HM peptide markers are bold. B) Venn diagram representing results of HCA of acquired mass spectra shows three primary cellular populations, A-type (n = 482), B-type (n = 386), and C-type (n = 211). The size of each set is drawn proportionally to reflect its total number of cells. “n” indicates the total number of cells in each population. C) Peptide markers are selected for cellular subgroups of A- type (LM m/z 4404), B-type (LM m/z 1947), and C-type (HM m/z 9437), and shared cells

(Overlap).

117

Mass Observed Peptide Primary sequence Exact m/z error m/z (ppm)

Neurofilament light Y.TSHVQEEQSEVEETIEATK.A 2174.007 2174.0095 -1.15 polypeptide

Neurofilament M.S(+42.01)YTLDSLGNPSAYRRV 2328.162 2328.1573 2.02 medium polypeptide TETR.S

Neurofilament light Y.YTSHVQEEQSEVEETIEATK.A 2337.079 2337.0729 2.61 polypeptide

Neurofilament light -- 2359.232a -- -- polypeptide

Neurofilament light Y.TSHVQEEQSEVEETIEATKAEE 2702.254 2702.2639 -3.66 polypeptide AK.D

L.LIKTVETRDGQVINETSQHHDD Vimentin 2777.365 2777.3700 -1.80 LE

Neurofilament light Y.YTSHVQEEQSEVEETIEATKAE 2865.317 2865.3272 -3.56 polypeptide EAK.D

M.A(+42.01)DKPDMGEIASFDKAK Thymosin beta-10 2964.491 2964.4977 2.26 LKKTETQEKN.T

Neurofilament K.VEAPKLKVQHKFVEEIIEETKV 3094.656 3094.6671 -3.59 medium polypeptide EDEK.S

G.VVRKEDLEPPAQDQEAKEQEE Gamma-synuclein 3297.528 3297.5353 -2.21 GEEAKSGGD

Unidentified -- 4131.338 -- --

Unidentified -- 4470.472 -- --

Unidentified -- 4404.672 -- --

Unidentified -- 4745.614 -- --

M.S(+42.01)DKPDMAEIEKFDKSK Thymosin beta-4 LKKTETQEKNPLPSKETIEQEKQA 4961.453 4961.4930 -8.06b GES Table 4.1 FT-ICR MS measurements of DRG tissue extracts. Peptide identity assignments are based on mass matching with the previous peptidomic characterizations using LC-MS.[41]

The “period” in the peptide sequences indicates the cleavage site; residues before and after it are not considered part of the primary sequence used to compute the m/z value. “+42”

118

Table 4.1 (cont.) indicates acetylation. The bolded m/z values were also detected in the single cell profiling experiment. a Signal observed in the prior LC-MS study, but the primary sequence of related compound was not determined. b A high ppm value suggests less confidence in the peptide match.

119

4.7 References

(1) J. V. Priestley, in Encyclopedia of Neuroscience (Ed.: L. R. Squire), Academic Press,

Oxford, 2009, pp. 935–943.

(2) S. N. Lawson, Morphological and Biochemical Cell Types of Sensory Neurons, Oxford

University Press, Oxford, 1992.

(3) K. L. Fields, J. P. Brockes, R. Mirsky, L. M. B. Wendon, Cell 1978, 14, 43-51.

(4) S. B. McMahon, J. V. Priestley, in The Neurobiology of Pain (Eds.: S. P. Hunt, M.

Koltzenburg), Oxford University Press, Oxford, 2005, pp. 35-64.

(5) I. M. Chiu, L. B. Barrett, E. Williams, D. E. Strochlic, S. Lee, A. D. Weyer, S. Lou, G.

Bryman, D. P. Roberson, N. Ghasemlou, C. Piccoli, E. Ahat, V. Wang, E. J. Cobos, C. L.

Stucky, Q. F. Ma, S. D. Liberles, C. J. Woolf, Elife 2014, 3.

(6) C. L. Li, K. C. Li, D. Wu, Y. Chen, H. Luo, J. R. Zhao, S. S. Wang, M. M. Sun, Y. J. Lu,

Y. Q. Zhong, X. Y. Hu, R. Hou, B. B. Zhou, L. Bao, H. S. Xiao, X. Zhang, Cell Res.

2016, 26, 967.

(7) D. Usoskin, A. Furlan, S. Islam, H. Abdo, P. Lonnerberg, D. Lou, J. Hjerling-Leffler, J.

Haeggstrom, O. Kharchenko, P. V. Kharchenko, S. Linnarsson, P. Ernfors, Nat.

Neurosci. 2015, 18, 145-153.

(8) S. S. Rubakhin, A. Ulanov, J. V. Sweedler, J. Am. Soc. Mass Spectrom. 2015, 26, 958-

966.

(9) S. S. Rubakhin, E. V. Romanova, P. Nemes, J. V. Sweedler, Nat. Methods 2011, 8, S20-

S29.

(10) A. Svatos, Anal. Chem. 2011, 83, 5037-5044.

120

(11) T. J. Comi, T. D. Do, S. S. Rubakhin, J. V. Sweedler, J. Am. Chem. Soc. 2017, 139, 3920-

3929.

(12) X. Chen, J. C. Love, N. E. Navin, L. Pachter, M. J. T. Stubbington, V. Svensson, J. V.

Sweedler, S. A. Teichmann, Nat. Biotechnol. 2016, 34, 1111-1118.

(13) S. S. Rubakhin, E. J. Lanni, J. V. Sweedler, Curr. Opin. Biotechnol. 2013, 24, 95-104.

(14) M. Qi, M. C. Philip, N. Yang, J. V. Sweedler, ACS Chem. Neurosci. 2017, 9, 40-50.

(15) H. M. Bergman, I. Lanekoff, Analyst 2017, 142, 3639-3647.

(16) Y. Yang, Y. Huang, J. Wu, N. Liu, J. Deng, T. Luan, TrAC, Trends Anal. Chem. 2017,

90, 14-26.

(17) H. Li, B. K. Smith, B. Shrestha, L. Mark, A. Vertes, Methods Mol. Biol. 2015, 1203, 117-

127.

(18) X. Gong, Y. Zhao, S. Cai, S. Fu, C. Yang, S. Zhang, X. Zhang, Anal. Chem. 2014, 86,

3809-3816.

(19) N. Pan, W. Rao, N. R. Kothapalli, R. Liu, A. W. Burgett, Z. Yang, Anal. Chem. 2014, 86,

9376-9380.

(20) J. K. Lee, E. T. Jansson, H. G. Nam, R. N. Zare, Anal. Chem. 2016, 88, 5453-5461.

(21) M. Stoeckli, P. Chaurand, D. E. Hallahan, R. M. Caprioli, Nat. Med. 2001, 7, 493-496.

(22) K. Schwamborn, R. M. Caprioli, Nat. Rev. Cancer 2010, 10, 639-646.

(23) A. Bodzon-Kulakowska, P. Suder, Mass Spectrom. Rev. 2016, 35, 147-169.

(24) B. Rocha, C. Ruiz-Romero, F. J. Blanco, Nat. Rev. Rheumatol. 2017, 13, 52-63.

(25) B. Spengler, Anal. Chem. 2015, 87, 64-82.

(26) I. Kaya, D. Brinet, W. Michno, M. Baskurt, H. Zetterberg, K. Blenow, J. Hanrieder, ACS

Chem. Neurosci. 2017, 8, 2778-2790.

121

(27) E. J. Lanni, S. S. Rubakhin, J. V. Sweedler, J. Proteomics 2012, 75, 5036-5051.

(28) C. Giesen, H. A. O. Wang, D. Schapiro, N. Zivanovic, A. Jacobs, B. Hattendorf, P. J.

Schuffler, D. Grolimund, J. M. Buhmann, S. Brandt, Z. Varga, P. J. Wild, D. Gunther, B.

Bodenmiller, Nat. Methods 2014, 11, 417-422.

(29) A. Zavalin, J. H. Yang, A. Haase, A. Holle, R. Caprioli, J. Am. Soc. Mass Spectrom.

2014, 25, 1079-1082.

(30) A. Zavalin, E. M. Todd, P. D. Rawhouser, J. H. Yang, J. L. Norris, R. M. Caprioli, J.

Mass Spectrom. 2012, 47, 1473-1481.

(31) A. Rompp, S. Guenther, Z. Takats, B. Spengler, Anal. Bioanal. Chem. 2011, 401, 65-73.

(32) M. Kompauer, S. Heiles, B. Spengler, Nat. Methods 2017, 14, 90-96.

(33) M. K. Passarelli, A. G. Ewing, Curr. Opin. Chem. Biol. 2013, 17, 854-859.

(34) P. L. Urban, K. Jefimovs, A. Amantonico, S. R. Fagerer, T. Schmid, S. Madler, J.

Puigmarti-Luis, N. Goedecke, R. Zenobi, Lab Chip 2010, 10, 3206-3209.

(35) A. J. Ibanez, S. R. Fagerer, A. M. Schmidt, P. L. Urban, K. Jefimovs, P. Geiger, R.

Dechant, M. Heinemann, R. Zenobi, Proc. Natl. Acad. Sci. U. S. A. 2013, 110, 8790-

8794.

(36) T. Fujii, S. Matsuda, M. L. Tejedor, T. Esaki, I. Sakane, H. Mizuno, N. Tsuyama, T.

Masujima, Nat. Protoc. 2015, 10, 1445-1456.

(37) T. H. Ong, D. J. Kissick, E. T. Jansson, T. J. Comi, E. V. Romanova, S. S. Rubakhin, J.

V. Sweedler, Anal. Chem. 2015, 87, 7036-7042.

(38) T. D. Do, T. J. Comi, S. J. B. Dunham, S. S. Rubakhin, J. V. Sweedler, Anal. Chem.

2017, 89, 3078-3086.

122

(39) T. J. Comi, M. A. Makurath, M. C. Philip, S. S. Rubakhin, J. V. Sweedler, Anal. Chem.

2017, 89, 7765-7772.

(40) E. T. Janson, T. J. Comi, S. S. Rubakhin, J. V. Sweedler, ACS Chem. Biol. 2016, 11,

2588-2595.

(41) E. G. Tillmaand, N. Yang, C. A. C. Kindt, E. V. Romanova, S. S. Rubakhin, J. V.

Sweedler, J. Am. Soc. Mass Spectrom. 2015, 26, 2051-2061.

(42) Q. Wu, J. L. Chu, S. S. Rubakhin, M. U. Gillette, J. V. Sweedler, Chem. Sci. 2017, 8,

3926-3938.

(43) A. P. Il'ina, O. G. Kulikova, D. I. Mal'tsev, M. S. Krasnov, E. Rybakova, V. S.

Skripnikova, E. S. Kuznetsova, A. K. Buriak, V. P. Iamskova, I. A. Iamskov, Prikl.

Biokhim. Mikrobiol. 2011, 47, 135-140.

(44) J. M. Boggs, Cell. Mol. Life Sci. 2006, 63, 1945-1961.

(45) J. M. Boggs, G. Rangaraj, Y. M. Heng, Y. F. Liu, G. Harauz, Biochim. Biophys. Acta

2011, 1808, 761-773.

(46) K. R. Brunden, Y. Ding, B. S. Hennington, J. Neurosci. Res. 1992, 32, 507-515.

(47) R. S. Nascimento, M. F. Santiago, S. A. Marques, S. Allodi, A. M. B. Martinez, Braz. J.

Med. Biol. Res. 2008, 41, 1011-1017.

(48) E. Perlson, S. Hanz, K. Ben-Yaakov, Y. Segal-Ruder, R. Seger, M. Fainzilber, Neuron

2005, 45, 715-726.

(49) V. Menet, M. Prieto, A. Privat, M. G. Y. Ribotta, Proc. Natl. Acad. Sci. U. S. A. 2003,

100, 8999-9004.

(50) J. Wong, M. M. Oblinger, J. Neurosci. Res. 1990, 27, 332-341.

(51) L. J. Boyne, I. Fischer, T. B. Shea, Int. J. Dev. Neurosci. 1996, 14, 739-748.

123

(52) M. Dubey, S. Hoda, W. K. H. Chan, A. Pimenta, D. D. Ortiz, T. B. Shea, J. Neurosci.

Res. 2004, 78, 245-249.

(53) M. J. Perry, S. N. Lawson, J. Robertson, J. Neurocytol. 1991, 20, 746-758.

(54) P. N. Hoffman, D. W. Cleveland, Proc. Natl. Acad. Sci. U. S. A. 1988, 85, 4530-4533.

(55) M. Yamamoto, T. Yamagishi, H. Yaginuma, K. Murakami, N. Ueno, Dev. Brain Res.

1994, 79, 177-185.

(56) T. J. Comi, E. K. Neumann, T. D. Do, J. V. Sweedler, J. Am. Soc. Mass Spectrom. 2017,

28, 1919-1928.

(57) N. Y. Agar, H. W. Yang, R. S. Carroll, P. M. Black, J. N. Agar, Anal. Chem. 2007, 79,

7416-7423.

(58) E. V. Romanova, S. S. Rubakhin, J. V. Sweedler, Anal. Chem. 2008, 80, 3379-3386.

(59) M. E. Goldstein, P. Grant, S. B. House, D. B. Henken, H. Gainer, Neuroscience 1996, 71,

243-258.

(60) C. Kilkenny, W. J. Browne, I. C. Cuthill, M. Emerson, D. G. Altman, PLoS Biol. 2010, 8,

e1000412.

(61) B. O. Keller, L. Li, J. Am. Soc. Mass Spectrom. 2001, 12, 1055-1063.

(62) B. O. Keller, L. Li, J. Am. Soc. Mass Spectrom. 2006, 17, 780-785.

124

CHAPTER 5

SINGLE CELL MALDI MS SUPERVISED BY IMMUNOCYTOCHEMICAL CLASSIFICATIONS

Notes and Acknowledgements

This chapter is adapted from a completed manuscript coauthored by Troy J. Comi,

Stanislav S. Rubakhin, and Jonathan V. Sweedler, submitted for publication on November 9,

2018, and accepted on February 27, 2019 (DOI: 10.1002/anie.201812892). S.S. Rubakhin performed aspects of sample preparation before MS analysis. T.J. Comi wrote a majority of the scripting in this work as well as wrote portions of the manuscript. J.V. Sweedler assisted with concept development and writing, as well as funded the work. We acknowledge Dr. Jennifer

Mitchell and Joseph F. Ellis for useful discussions. The authors gratefully acknowledge support from the National Institutes of Health, Award Number P30 DA018310 from the National

Institute on Drug Abuse, and from the National Institute of Mental Health, Award Number 1U01

MH109062. E.K. Neumann and T.J. Comi acknowledge support from the National Science

Foundation Graduate Research Fellowship Program and the Springborn Fellowship. T.J. Comi received additional support through the Training Program at Chemistry-Interface with Biology

(T32 GM070421).

5.1 Introduction

Mammalian brain function critically relies upon cellular heterogeneity,[1-3] and explorations of this heterogeneity have exploded with the advent of single cell transcriptomics.[4]

Cell morphology, the most traditional metric of cell classification, correlates shape to the function of cells.[5,6] Cell chemistry also correlates to function; for example, the biophysical properties of the cell membrane are governed by its lipid composition.[7-9] Chemical analysis of single cells has entered a new era because of recent advances in mass spectrometry (MS).[10] For

125 example, high-throughput single cell matrix-assisted laser desorption/ionization (MALDI) MS is capable of directly assaying hundreds of chemical compounds within biological specimens.[11]

While single cell MALDI MS produces rich spectral profiles for label-free classifications, the results can be difficult to correlate to canonical cell types.[12, 13]

Immunocytochemistry (ICC) is effective for classifying cells as it can be used to target a plethora of different molecules, such as neurotransmitters, neuropeptides, and proteins that distinguish cell types.[14] Fluorescence overlap of labels typically limits plexity to fewer than ten compounds.[15] Most ICC protocols require sample fixation and lipid removal, crosslinking and/or removing many compounds that are easily ionized by MALDI MS.[16] ICC has been combined with tissue imaging and a few examples applied the methods to single cell MS.[17, 18]

Directly coupling ICC and MS on the same cell was reported in 2012,[18] however, ICC was performed first, limiting the analysis to high-abundance peptides within a small number of cells.

Here we describe a workflow enabling high throughput MALDI MS of thousands of cells, followed by cell classification via ICC. By performing ICC after MALDI-MS, we can expand chemical interrogation to include lipids and small metabolites, which are essential for brain function.[19-22] Our dataset here consists of over 1800 rodent cerebellar cells assayed using microMS, a cell finding and analysis software.[23] We demonstrate the approach by examining chemical differences between two of the most common cell types in the brain – astrocytes and neurons. Because the lipid differences between astrocytes and neurons are both poorly understood and modest, ICC classification was imperative to discovering statistically distinct features between these important cell types, particularly within primary isolated and mixed cell populations.

126

5.2 Experimental

Chemicals.

2,5-Dihydroxybenzoic acid (DHB), bovine serum albumin (BSA), Tween 20, and ethanol were purchased from Sigma-Aldrich (St. Louis, MO). 4% paraformaldehyde (PFA) in phosphate buffered saline (PBS) was purchased from Affymetrix (Santa Clara, CA). Anti-glial fibrillary acidic protein (mouse anti-GFAP) (MAB360) and anti-neurofilament light chain (rabbit anti-NF-

L) (AB9568) were purchased from EMD Millipore (Billerica, MA). Dylight 550 secondary antibody (goat anti-mouse) was purchased from Thermo Scientific (Hudson, NH). Alex Fluor

488 secondary antibody (goat anti-rabbit), phosphate buffered saline, Hoechst 33342, and 10% normal goat serum were purchased from Life Technologies (Gaithersburg, MD). Peptide

Calibration Standard Kit II (angiotensin II, angiotensin I, substance P, bombesin, ACTH clip

1−17, ACTH clip 18−39, somatostatin 28, bradykinin fragment 1−7) was purchased from Bruker

Corp. (Billerica, MA). All other reagents were purchased from Sigma Aldrich. All reagents were used as received (> 98% purity) without further purification.

Animals.

Six, 2–2.5-month-old male Sprague Dawley outbred rats (Rattus norvegicus)

(www.envigo.com) were housed on a 12-h light cycle and fed ad libitum. Animal euthanasia was performed in accordance with the appropriate institutional animal care guidelines (the

Illinois Institutional Animal Care and Use Committee), and in full compliance with federal guidelines for the humane care and treatment of animals.

Cell Dissociation.

Dissected cerebellum tissues were incubated with a solution of 1% Hoechst 33342 in oxygenated modified Gey’s balanced salt solution (mGBSS) for 30 minutes at 37°C. The

127 mGBSS solution was removed and tissues incubated with an oxygenated solution of 6 units/mL of papain, 1 mM L-cysteine, and 0.5 mM ethylenediaminetetraacetic acid for 80 minutes at 37°C.

Tissue was then mechanically dissociated in mGBSS with 0.04% PFA for cell stabilization through mechanical dissociation. The PFA concentration was optimized to maximize cell integrity and quality of mass spectra (data not shown). Following dissociation, glycerol was added to a final concentration of 40% and 150 µL of cell suspension was transferred onto indium tin oxide (ITO) coated glass slides (Delta Technologies, Loveland, CO) or non-coated glass slides (LabMed, Ft. Lauderale, Florida) for MS analysis or immunostaining control experiments, respectively. Fiducial markers used for registration were etched using a diamond-tipped pen prior to cell plating.

Each slide held cells from three biological replicates plated in duplicate in a random coordinate on a gridded pattern. This design of experiments helps control batch variation and possible position-dependent artifacts. Immunostaining controls, consisting of a mixture of three animals, were placed on diagonals at the upper right and lower left positions the slide (Figure

5.1). Six animals were used for a total of eight ITO and eight glass slides. Cell populations were kept isolated with stretched parafilm “M” strips while the cells were adhering (Bemis Company,

Inc., Neenah, WI). After ~18 hours, the parafilm “M” strips were removed and cells were stained with Hoechst 33342 (1 µg/mL in mGBSS) for 15 minutes before quickly rinsing twice with 500

µL of 150 mM ammonium acetate.

Optical Imaging.

Brightfield and fluorescence images were acquired on a Zeiss Axio M2 microscope

(Zeiss, Jena, Germany) equipped with an Ab cam Icc5 camera, X-cite Series 120 Q mercury lamp (Lumen Dynamics, Mississauga, Canada), and a HAL 100 halogen illuminator (Zeiss, Jena,

128

Germany). DAPI (ex. 335-383 nm; em. 420-470 nm), GFP (ex. 450-490 nm; em. 500-550), and

DsRed (ex. 538-562 nm; em. 570-640) dichroics were used for fluorescence excitation. The images were acquired with a 10x objective (1 pixel width is 0.55 µm) with a 13% overlap produced during image tiling. Images were processed and exported on ZEN software version 2 blue edition (Zeiss, Jena, Germany). Images were exported as big tiff files. Only Hoechst 33342 fluorescence images were acquired for MALDI-MS analysis, whereas all three filters were used after ICC staining. Cells without antibody incubation were used to measure cellular autofluorescence during image post-processing. microMS Analysis.

microMS was used as previously described. Briefly, cells were found using the high fluorescence from Hoechst 33342 and were filtered by size, shape and distance to reduce time spent analyzing debris and artifacts from cell sampling. microMS was used to register microscopy images with the MALDI MS stage.

Matrix Application and MS Analysis.

After full-slide imaging was performed, slides were coated with a 50 mg/mL DHB solution dissolved in 1:1 ethanol: water with 0.1% trifluoroacetic acid using an automatic sprayer

[39] described previously. The matrix solution was nebulized at 10 mL/hr using N2 gas at 50 psi with 100 passes. Samples were taped to a rotating plate and the spray was placed 3 cm above the samples. The total amount of matrix applied was between 0.1 and 0.2 mg/cm2.

Single-cell analysis was performed on an ultrafleXtreme II TOF/TOF mass spectrometer

(Bruker Corp., Billerica, MA) with a mass window of 500-3000. The “Ultra” (~100 µm footprint) laser setting was used and 300 laser shots were accumulated at 1000 Hz and 60% laser

129 energy for each cell. Cell coordinates were obtained in microMS and filtered with a 100 µm distance filter to prevent cellular contamination.[41] (Figure 5.2A).

Cerebral lipid extracts were prepared using the Bligh-Dyer method. Direct infusion electrospray ionization of a cerebellum lipid extract (~1 mg of dried cerebral lipid extract was resuspended in 1 mL of 50:50 methanol and water) was performed for high mass accuracy and collision induced dissociation (CID) of selected lipids. Lipid identifications were performed with a solariX XR 7T Fourier transform ion cyclotron resonance (FT-ICR) mass spectrometer (Bruker

Corp., Billerica, MA) with a mass window of either 100-3000 or 150-3000, yielding a transient length of 1.96 and 2.94 seconds, respectively. Direct infusion electrospray ionization (ESI) of a cerebellum lipid extract was performed for high mass accuracy and collision induced dissociation of lipids (Figure 5.3). Sample was delivered at 120 µl/hr and a capillary voltage of

3900 V. Complete isolation of the lipids was sometimes not possible, so there may be contamination of tandem MS spectra by neighboring lipid features; however, the most abundant fragment ions were used for feature assignment. The LIPID MAPS database was used in tandem with CID spectra for assigning lipid species.[1] CID spectra were recalibrated using the molecular ion.

Matrix Removal, Sample Fixation, and ICC.

Matrix was removed during cellular fixation by incubating the cells in a 4% solution of

PFA in PBS for five minutes. The fixation solution was then changed and the cells were incubated for an additional ten minutes followed by a three-minute rinse in PBS. Slides were then incubated for 45 minutes at room temperature in a blocking buffer solution consisting of 50 mg/mL BSA, 10% (v/v) goat serum, and 0.5% (v/v) Tween-20 in PBS. Slides were rinsed again in PBS for three minutes following blocking. A hydrophobic pen was used to isolate antibody

130 controls and reduce the quantity of antibody solution required. Fixed cells were immunostained with primary antibodies against GFAP and NF-L for two days at 4°C in antibody buffer (1%

BSA and 0.5% Tween-20 in PBS; 1:1000 dilution of each antibody). Cells were then stained over another two-day incubation with the fluorescently labelled secondary antibodies in the same antibody buffer at 1:100 dilution (Figure 5.2B). Slides were washed in PBS three times for three minutes after each antibody incubation. Stained samples were imaged using an Axio M2 imager as described above (Figure 5.2C).

Image Registration and Analysis.

To relate the immunostained images to the mass spectra, the cell locations were identified by registering the images obtained before MS with the post-ICC images using the custom

MATLAB (R2015B) scripts provided as Supplementary Information (WinZip file). First, brightfield images taken before matrix application and after immunostaining underwent point- based registration via manual overlay, which greatly improved the performance of later, automatic registration procedures. For each cell analyzed by MALDI MS, a 90-pixel by 90-pixel area with the cell in the center was located in the post-ICC image using the initial, manual transformation. An intensity-based registration was then performed between the two Hoechst

33342 fluorescence images, which highlight the cell nucleus. Cells that did not display fluorescence signal after matrix removal or failed to converge upon registration were eliminated from analysis. The removed population represented cells that moved further than ~25 µm from the starting point or changed drastically following staining (e.g., because of contamination from debris or other cells). It was necessary to perform automatic registration on a cell-by-cell basis to correct any stitching artifacts and ensure accurate overlays of cell positions. Red and green intensities within the ICC image were averaged at each cell location over an area corresponding

131 to the cell boundary, as determined by microMS.[4] Following this procedure, each cell had a corresponding mass spectrum and average fluorescence intensity in each channel.

The fluorescence intensities were used to classify the immunoprofiles with batch-specific cutoffs. Two fluorescence controls (antibody blanks) located on the opposite corners of the slide were used to determine the intensity values for a positive fluorescence signal. A histogram of fluorescence intensity was produced, and an appropriate cutoff determined as a constant value

(5% of max range) above the maximum fluorescence intensity of the cells located within the antibody blanks.

The ICC-enriched dataset was utilized for supervised multivariate analysis of the MALDI

MS results. Single cell mass spectra were imported using the readbrukermaldi

(https://github.com/AlexHenderson/readbrukermaldi) function, binned at 0.05 Da for peak peaking and normalized to total ion count (TIC). Spectra were sorted into GFAP-positive and

NF-L-positive groups; cells with neither stain were excluded because a negative stain could be attributed to methodological artifacts.

Statistics.

Comparisons between cells that were immediately fixed and cells that underwent MALDI matrix removal and subsequent ICC were compared using a two-tailed t-test (n = 6 animals, p- value = 0.002) because the staining profiles were normally distributed. The total number of stained cells was statistically different at the 95% confidence interval. Comparisons of staining proportions between matrix-removed and control ICC were made using the proportions test (ICC only, number of cells = 5435; matrix removal, number of cells = 5646; n = 6 animals for both treatments, p-value = 0.886 and z-value = –0.49). A proportions test was used to compare the proportion of stained cells under each condition, rather than the average number in each

132 population. A rank sum test was used to determine ions of significant difference between the two immunostained populations with false discovery rate correction (n = 6 animals). A rank sum test was used to determine significant metabolic differences between the two cell classes as opposed to a t-test because the MS data was not normally distributed. As such, a rank sum test is the most appropriate statistical test to use, because it does not depend on normality. T-distributed stochastic neighbor embedding and hierarchical clustering were performed using default

MATLAB functions to visualize heterogeneity possessed by NF-L and GFAP positive cells.

Optimal clusters were determined via the evalclusters function utilizing silhouette evaluation.

Number of Cells at Each Stage.

~586,000 structures were deposited on eight slides and, of these, ~275,000 were ICC positive. To reduce cell-to-cell contamination during MS analysis, only fluorescent, cell-like structures that were >100 µm away (two-fold greater than the diameter of the laser probe) from other cell-like structures were analyzed by MALDI MS (~41,000 structures). Of these 41,000 structures, 1816 cells had detectable lipids, were not delocalized during the matrix removal or

ICC processes, not covered with the hydrophobic pen, and were ICC positive. Separately, the number of ICC-positive cells for statistical comparison of ICC-only (5435 cells) and MALDI matrix-removed (5646 cells) conditions were manually determined because microMS did not allow us to distinguish bare nuclei without a cellular membrane. Sixty areas of ~100 cells of each condition over the eight slides were used to count the number of ICC-positive cells, ICC- negative cells, and bare nuclei. Bare nuclei were disregarded in the statistical comparisons.

5.3 Results and Discussion

Most ICC procedures greatly reduce MS quality, especially from single cells where material is limited. While mass spectra acquired from unfixed single cells contained peaks

133 consistent with small metabolites and lipid species at high intensity, after performing the fixation procedure (as required for most ICC protocols), the resulting mass spectra showed weaker metabolite peak intensities (~90% lower) as well a polymer contaminant (Figure 5.1). Thus, we decided to perform ICC after MALDI MS (Figure 5.1A-C) similarly to our recent paper coupling vibrational spectroscopy and MALDI MS;[24] using MALDI MS first, we observe high-quality mass spectra of molecules, including small metabolites and lipids. For this procedure to succeed, next we need to remove the MALDI matrix (e.g., 2,5-dihydroxybenzoic acid) that is thoroughly co-crystalized with the cells, keep the cells anchored to the same location on the slide, and finally, the antibody must still recognize its epitope, even after the MALDI measurement. We found that the common fixative paraformaldehyde was effective in both fixing the cells and removing the MALDI matrix (Figure 5.3G). Importantly, a high proportion of cells (~95%) did not move during matrix removal (Figure 5.4, insets). While ICC requires long incubations that can lead to cell delocalization, we find that between matrix removal (Figure 5.4A, left) and second antibody incubation (Figure 4A, right), ~88% of the cells stayed in the same location, while only a few were delocalized (Figure 4A, red arrows). Relying on image registration as a metric for cell retention, the registration procedure used here successfully matched 71±18% of cells within a 25 µm field of view.

Using the optimized procedure, cells within the cerebellum were classified as glial fibrillary acidic protein (GFAP) positive, neurofilament-L (NF-L) positive, or ICC negative.

While GFAP is not a universal astrocytic marker, most astrocytes within the cerebellum can be labeled immunopositively by GFAP. Similarly, most neurons express NF-L antigens. The cells that were ICC negative may be other cell types that lack these antigens, or they may be cells damaged during cell dissociation and/or MALDI MS process, and thus did not immunostain

134 appropriately. Due to their ambiguous identification, ICC-negative cells were excluded from the mass spectral analysis.

Ultimately, ~586,000 cell-like structures were deposited on eight slides. To reduce cell- to-cell contamination during MS analysis, only fluorescent, cell-like structures that were >100

µm away (two-fold greater than the diameter of the laser probe) from other cell-like structures were analyzed by MALDI MS (~41,000 structures). Of these 41,000 cells, 10,000 were in the control areas on the slide that were not immunostained. Further, ~12,000 cells were not present within the 25 µm field of view and were assumed to be delocalized. Bare nuclei were not excluded from the analysis because Hoechst 33342 is signal is indistinguishable between these two populations. We assume that bare nuclei are not ICC positive and negatively impact the

MALDI-MS ICC efficiency. Of the remaining cells, 1816 cells were ICC positive. It is imperative for chemical analysis of thousands of cells to be feasible because of the brain’s complexity, which we have achieved here.

Previous lipidomic research of astrocytes and neurons relied on cell lines or cultured primary cells to study metabolic differences.[25-27] The chemical nature of neurons and astrocytes results from their position within a large network of cells; all contextual information is lost during cell culture. Astrocyte cell lines for example, lose GFAP immunoreactivity, the primary astrocytic marker. Because the dissociated cells used here are rapidly transferred from the brain, they retain many defining features, albeit the native cell position is entirely lost. The chemical information gained from dissociated cells is similar to primary astrocyte cultures.[28] Our protocol, however, allows for targeted analysis of cell types, without having to purify cultures of a specific cell type and could easily be adaptable to primary cultures.

135

To assess the effect of MALDI MS on the ICC profiles, immunostaining efficiency was compared between cells subjected to matrix removal and cells that were fixed and immunostained immediately after dissociation. Unsurprisingly, cells stained more efficiently when immediately fixed and stained (Figure 5.4B). The differences in staining could be caused by many factors, including the effects of matrix removal or laser irradiation. On control slides,

72% of cells stained, with this number decreasing to 46% following matrix removal, after bare nuclei were removed from cell counts. The discrepancy between treatments is significant (p- value = 0.002, two-tailed t-test), but if we assume the control treatment as an upper limit, more than half of the cells that could immunostain were still characterized following MALDI MS.

Additionally, the relative amount of each population remained consistent between control and matrix-removed cells (p-value = 0.886 and z-value = –0.49; proportions test). The consistent proportion of GFAP- and NF-L-positive cells is important; if one population had been disproportionately affected by the MALDI MS measurement process, the derived chemical inferences may be affected.

Using our MALDI MS-ICC approach, we can directly correlate the ICC profile, cell image, and mass spectrum for each cell (Figure 5.5A-D). Both NF-L positive cells (Figure 5.5A-

B) and GFAP-positive cells (Figure 5.5C-D) have the capacity to stain with high-quality mass spectra that include many lipid features that are absent from MS of fixed cells. Even after laser ablation and matrix removal, each cell remains intact (Figure 5.5A-D insets) and maintains structural features present during cell plating for easy recognition post-analysis, except for what appears to be nuclei leakage that sometimes occurs after MS processing (Figure 5.5C). Even though the cell is visibly damaged, it still immunostains, demonstrating the robustness of the targeted antigens. While cells may stain similarly, they may not possess identical mass spectral

136 features revealing the importance of single cell measurements as well as bulk measurements.

Further, the heterogeneity of each class of cells can further be visualized by t-distributed stochastic neighbor embedding (t-SNE) and hierarchical clustering analysis (HCA) (Figure 5.5E-

F). NF-L and GFAP positive cells can be subsequently clustered into five and four subpopulations, respectively, based on their mass spectral features. The presence of multiple clusters may demonstrate that even though cells may stain similarly, they may not possess identical mass spectral features and could belong to subcellular populations. Thus, our preliminary results revealing the importance of single cell measurements as well as bulk measurements for chemical analysis of the brain by MS.

While lipids are known to have important biological functions in the brain relating to communication, metabolism, and transport, the specific function of many individual lipids is unknown. Partly, this results from the difficulty in measuring large numbers of lipids in complex samples, identifying the saturation points of lipids, and limited databases containing lipidomic information (particularly when compared to genomics and proteomics).[29] MS is one of the only techniques capable of lipid identification in complex samples, such as single cells, and is one of the leading methods for lipidomic analysis. Because MALDI MS is a highly sensitive technique, we detected an average of 99 molecular features within the single cells and found that 17 lipids had statistically distinct abundances between neurons and astrocytes (Table 5.1). Statistical differences were determined using a rank sum test with a false discovery rate correction and high-resolution FT-ICR MS analysis was used for identification. Ultimately, we were unable to perform direct tandem MS of the lipids from single cells that were ICC positive because of the required fixation step. As such, we performed a lipid extract of bulk cerebral tissue for high mass accuracy and tandem MS FT-ICR MS analysis to assign the statistically significant lipids under

137 these experimental conditions. The lipids of interest were detected within multiple cells, so we believe that they would also be detected in bulk tissue extract. While there are several examples of multiple lipids resolved by FT-ICR MS analysis making assignments ambiguous (Table 5.1, ex. m/z value 731), many of the lipids were sufficiently isolated from other spectral features, increasing confidence of the lipid assignments. All but three lipids had mass errors of less than 1 ppm and tandem MS (MS/MS) was possible for all but four of the lipids with errors less than 5 ppm (Table 5.1).

The 17 lipid differences that we detect here would likely not be observed in bulk tissue analysis because of ion suppression, instrumental dynamic range, and difficulties isolating neurons and astrocytes. Because we perform single cell analysis and can average groups of cells together, we can identify small, yet consistent lipid differences between astrocytes and neurons.

Biological membranes are highly conserved because they must spontaneously form lipid bilayers within water, so differences may relate to essential functional or structural. Determining the function of these lipids is beyond the scope of this communication; however, combining MALDI

MS with ICC has narrowed targets for functional analyses down to a small number of lipids from the thousands known to be present in the brain.

From lipidomics, the functional consequences of the lipid differences can be putatively explored. Five lipids were present at higher intensity within astrocytes, four of which are phosphatidylethanolamines (PEs), such as PE(36:1). PE lipids, in particular, are important for determining membrane curvature as well as contractile ring disassembly at the cleavage furrow following cytokinesis.[30-33] The higher PE abundance in astrocytes may reflect their replication rate compared to neurons and their greater curvature. PE lipids have also been implicated in vesicle fusion and release for neurotransmission, potentially explaining the presence of PEs in

138 neurons as well as astrocytes.[33, 34] Another 12 lipids had higher abundances in neurons (e.g.

PC(40:6) and TG(42:4). Many of the statistically different lipids enriched within neurons include phosphatidylcholines (PCs), which are common in lipid membranes and serve as the source for choline used in neurotransmission.[35] The remaining lipids are more difficult to relate to cell physiology; however, their relative abundance may reflect the distinct shapes or functions of neurons or astrocytes. Because biological membranes are highly conserved structures, we are excited that single cell MALDI MS can detect these subtle differences and ICC allows us to connect them to cell type.

5.4 Conclusions

While recent single cell heterogeneity studies within brain cells have primarily been spearheaded by transcriptomics, these studies are inherently devoid of information on the metabolites physically present within the cell. Because of this gap in information, we have developed an additional avenue for determining single cell metabolic information that can be tied directly to canonical cell types as defined by ICC. Not only can functional information be garnered from metabolites present within the cell, but intrinsic chemical differences, such as lipid composition, may provide a more robust classification metric with direct links to cell phenotypes. The information gained from MALDI MS-ICC can guide targeted transcriptomic experiments to further our knowledge of how subtle chemical changes impact the global function of the brain. It also enables targeted functional lipid experiments that are imperative for understanding many disorders as well as normal brain function. Moreover, the precise mapping of chemical information with ICC-established cell types within the brain is a significant advancement in single cell MS as previous reports have determined single cell spectral profiles for cell lines or from a few specific cells isolated from a brain region. Our development allows

139 the merging of the rich chemical information found within mass spectra to the well-defined cellular classifications from ICC from thousands of cells, enabling new insights into cell-to-cell differences found within the cells of the brain.

5.5 Acknowledgements

The authors gratefully acknowledge support from the National Institutes of Health,

Award Numbers 1U01 MH109062 from the National Institute of Mental Health and P30

DA018310 from the National Institute on Drug Abuse. E.K.N. and T.J.C acknowledge support from the National Science Foundation Graduate Research Fellowship Program and Springborn

Fellowships. The authors would also like to thank Jennifer Mitchell for useful discussions.

140

5.6 Figures and Table

Figure 5.1 A) Representative single cell spectrum from an unfixed cerebellum cell. B) Single cell spectrum of the same cell from A following fixation. After fixation, the MALDI-MS spectrum contains a number of artifacts and additional features that hinder characterization of cellular content.

141

Figure 5.2 Schematic representation of the major steps of the MALDI MS-ICC approach. A) microMS is used to find the location of the ~586,000 cells on the slides. Single cell MALDI MS is performed on ~41,000 cell-like structures that are 100 µm apart. B) Simultaneous MALDI matrix removal and cell fixation with a solution of 4% PFA in PBS was performed before immunostaining against GFAP and NF-L. ~275,000 cells on the slide immunostain successfully.

C) Mass spectra and ICC profiles can be directly compared because cells stay in the same location. 1816 cells contained quality mass spectra and ICC profiles.

142

Figure 5.3 Cells remain in the same position after MALDI matrix removal and cell fixation if an organic solution is not present but are dislocated in other cases. Overlaid Hoechst-stained images are used to track cell positions, where purple represents high intensity Hoechst fluorescence (the position of a cell) in the initial image and green represents high intensity Hoechst fluorescence in the image following treatment. A) Overlaid image of a slide after matrix removal using ethanol, where most of the cells are delocalized. B, C) Magnified regions of cells pre- (top) and post-

143

Figure 5.3 (cont.) treatment (bottom). D) Using an aqueous solution, cells remain in the same position during cell fixation (before (top) and after (bottom) fixation). E, F) Magnified regions of cells demonstrating they retain their orientation. G) Overlaid image of a slide with simultaneous matrix removal and fixation with H, I magnified regions showing the cells stay in the same position after matrix removal (before (top) and after (bottom) fixation). The scale bar is 200 µm

(A, D, G) and 50 µm (B, C, E, F, H, I).

144

Figure 5.4 A) The left image is a subset of cells before the ICC procedure and the right image is a subset of cells after ICC (scale bar = 100 µm). Red arrows indicate several cells that were dislocated during the ICC procedure. B) Graph comparing a subset of immunopositive cells in dissociated cells that were immediately fixed (ICC only, number of distinct cells = 5435, n = 6 animals) and cells that underwent MALDI matrix removal and subsequent ICC (number of distinct cells = 5646, n = 6 animals). The total number of stained cells were statistically different at the 95% confidence interval and error bars represent the standard deviation. The proportion of

GFAP- to NF-L positive cells was not statistically different between the matrix-removed and

ICC-only cells.

145

Figure 5.5 High-quality single cell mass spectra and ICC images can acquired and compared for over a thousand cells. A) Single cell spectrum for the NF-L-positive cell displayed in the inset.

B) Second NF-L positive cell with similar mass spectral features. C) GFAP-positive single cell

146

Figure 5.5 (cont.) spectra that is visibly different from the second chosen GFAP-positive cell

(D). Green and red fluorescence corresponds to NF-F and GFAP, respectively, while blue fluorecsence corresponds to a nuclear dye. E) t-SNE plot of NF-L positive cells colored to correspond to the five clusters determined by HCA. F) t-SNE plot of GFAP positive cells colored to correspond to the four clusters determined by HCA.

147

Figure 5.6 A) Example slide layout for MALDI-MS with follow-up ICC. The “ICC control” areas contain a mixture of cells from all three animals. ICC controls are subjected to laser ablation but not antibody incubation with either primary or secondary antibody and help establish fluorescence intensity cutoffs during image analysis. Numbered regions indicate the location of three animals on the slide, with duplicates. The position of cells from each animal and its replicate are randomized while ICC controls are always positioned on the diagonals. B)

Parafilm M is added to prevent cells from intermixing during settling.

148

Figure 5.7 Collision induced dissociation (CID) spectra for compounds assigned in Table 5.1.

Assignments were made based on characteristic fragments (e.g., PC head group at m/z

184.0733), the LIPID MAPS database, and high mass accuracy measurements (<5 ppm).

Putative lipid species were identified using LIPID MAPS and checked against CID spectra for the presence of the head group and side chain fragments. Positive assignments were made if characteristic head group and lipid tails were manually identified within the CID spectra with

149

Figure 5.7 (cont.) less than a 5 ppm mass error. CID energies were varied for optimal signal-to- noise ratios of fragment ions.

150

Relative Measured m/z Value PPM Error Assignment Type Enrichment GFAP 837.6607‡ Unknown PS Lipid 61% Positive [PE(O-34:2)+Na]+ GFAP 724.5249 -0.3602 58% [PE(P-34:1)+Na]+ Positive GFAP 789.6143* 3.3624 [PE(38:2)+NH ]+ 39% 4 Positive [PE(O-36:2)+Na]+ GFAP 752.5564 -0.0810 28% [PE(P-36:1)+Na]+ Positive 768.5514 [PC(O-34:1)+Na]+ 0.0312 GFAP 16% 0.3110 Positive 768.5880 [PC(P-34:0)+Na]+ NF-L 850.6730‡ unknown 60% Positive 731.4534‡ NF-L -0.2064 [SM(d36:1)+H]+ 30% Positive 731.6060‡ NF-L 732.5538 0.0246 [PC(32:1)+H ]+ 27% Positive NF-L 834.6783‡ 1.2460 [HexCer(d42:1)+Na]+ 26% Positive NF-L 784.5828* 0.1580 [PC(34:0)+Na]+ 24% Positive NF-L 828.5504 -1.1780 [PC(38:6)+Na]+ 23% Positive NF-L 756.5513 -0.1004 [PC(32:0)+Na]+ 22% Positive NF-L 808.5825* -0.2177 [PC(36:2)+Na]+ 21% Positive NF-L 782.5670 -0.0332 [PC(34:1)+Na]+ 18% Positive NF-L 815.6347‡ unknown 16% Positive 762.5044* [PE(36:4)+Na]+ -0.0341 NF-L 11% -0.4353 Positive 762.6004* [PC(34:0)+H]+ 772.5253‡ [PC(32:0)+K]+ 0.0168 NF-L 06% 0.4310 Positive 772.6218‡ [CerP(t44:2)+H]+

Table 5.1 (cont.) List of compounds that were determined to be significantly different (p- value<<0.05) between GFAP- and NF-L-positive cells (n = 1547). Compounds were identified

151

Table 5.1 (cont.) using cerebellum lipid extracts with 7T Fourier transform ion cyclotron resonance mass spectrometry via high mass accuracy and tandem MS unless otherwise indicated.

152

5.7 References

(1) Q. Li, B. A. Barres, Nat. Rev. Immunol. 2017, 18, 225.

(2) M. Morales, E. B. Margolis, Nat. Rev. Neurosci. 2017, 18, 73.

(3) C. Schitine, L. Nogaroli, M. R. Costa, C. Hedin-Pereira, Front. Cell. Neurosci. 2015, 9.

(4) J.-F. Poulin, B. Tasic, J. Hjerling-Leffler, J. M. Trimarchi, R. Awatramani, Nature

Neuroscience 2016, 19, 1131.

(5) M. Uusisaari, K. Obata, T. Knöpfel, Journal of Neurophysiology 2007, 97, 901-911.

(6) J. Grosche, H. Kettenmann, A. Reichenbach, Journal of Neuroscience Research 2002,

68, 138-149.

(7) K. Simons, W. L. C. Vaz, Annual Review of Biophysics and Biomolecular Structure

2004, 33, 269-295.

(8) G. van Meer, D. R. Voelker, G. W. Feigenson, Nat. Rev. Mol. Cell Biol. 2008, 9, 112.

[9] P. A. Janmey, P. K. J. Kinnunen, Trends in Cell Biology 2006, 16, 538-546.

(10) X. Chen, J. Christopher Love, N. Navin, L. Pachter, M. J T Stubbington, V. Svensson, J.

V Sweedler, S. A Teichmann, Single-cell analysis at the threshold, Vol. 34, 2016.

(11) D. S. Cornett, M. L. Reyzer, P. Chaurand, R. M. Caprioli, Nat. Methods 2007, 4, 828.

(12) T. D. Do, T. J. Comi, S. J. B. Dunham, S. S. Rubakhin, J. V. Sweedler, Analytical

Chemistry 2017, 89, 3078-3086.

(13) S. S. Rubakhin, E. V. Romanova, P. Nemes, J. V. Sweedler, Nature methods 2011, 8,

S20-29.

(14) F. Doetsch, I. Caillé, D. A. Lim, J. M. García-Verdugo, A. Alvarez-Buylla, Cell 1999, 97,

703-716.

(15) E. C. Stack, C. Wang, K. A. Roman, C. C. Hoyt, Methods 2014, 70, 46-58.

153

(16) W. J. Howat, B. A. Wilson, Methods 2014, 70, 12-19.

(17) L. Li, E. V. Romanova, S. S. Rubakhin, V. Alexeeva, K. R. Weiss, F. S. Vilim, J. V.

Sweedler, Analytical Chemistry 2000, 72, 3867-3874.

(18) S. Neupert, S. S. Rubakhin, J. V. Sweedler, Chem. Biol. 2012, 19, 1010-1019.

(19) D. Piomelli, O. Sasso, Nature Neuroscience 2014, 17, 164.

(20) R. J. A. Wanders, B. T. Poll-The, Neuroscience Letters 2017, 637, 11-17.

(21) R. P. Bazinet, S. Layé, Nature Reviews Neuroscience 2014, 15, 771.

(22) C. Leloup, C. Allard, L. Carneiro, X. Fioramonti, S. Collins, L. Pénicaud, Neuroscience

2016, 323, 110-120.

(23) T. J. Comi, E. K. Neumann, T. D. Do, J. V. Sweedler, Journal of The American Society

for Mass Spectrometry 2017, 28, 1919-1928.

(24) E. K. Neumann, T. J. Comi, N. Spegazzini, J. W. Mitchell, S. S. Rubakhin, M. U.

Gillette, R. Bhargava, J. V. Sweedler, Analytical Chemistry 2018.

(25) K. Nieweg, H. Schaller, F. W. Pfrieger, Journal of Neurochemistry 2009, 109, 125-134.

(26) S. A. Moore, E. Yoder, S. Murphy, G. R. Dutton, A. A. Spector, Journal of

Neurochemistry 1991, 56, 518-524.

(27) F. W. Pfrieger, Cellular and Molecular Life Sciences CMLS 2003, 60, 1158-1171.

(28) S. C. Lange, L. K. Bak, H. S. Waagepetersen, A. Schousboe, M. D. Norenberg,

Neurochemical research 2012, 37, 2569-2588.

(29) M. R. Wenk, Nature Reviews Drug Discovery 2005, 4, 594.

(30) J. E. Vance, G. Tasseva, Biochimica et Biophysica Acta (BBA) - Molecular and Cell

Biology of Lipids 2013, 1831, 543-554.

154

(31) K. Emoto, T. Kobayashi, A. Yamaji, H. Aizawa, I. Yahara, K. Inoue, M. Umeda,

Proceedings of the National Academy of Sciences of the United States of America 1996,

93, 12867-12872.

(32) K. Emoto, M. Umeda, in Regulation of Contractile Ring Disassembly by Redistribution

of Phosphatidylethanolamine, Vol. 149, 2000, pp. 1215-1224.

(33) P. E. Glaser, R. W. Gross, Biochemistry 1995, 34, 12193-12203.

(34) K. Lohner, Chemistry and Physics of Lipids 1996, 81, 167-184.

(35) T. J. Tracey, F. J. Steyn, E. J. Wolvetang, S. T. Ngo, Frontiers in Molecular

Neuroscience 2018, 11, 10.

155

CHAPTER 6

LIPID ANALYSIS OF THIRTY-THOUSAND INDIVIDUAL RODENT CELLS USING HIGH-

RESOLUTION MASS SPECTROMETRY

Notes and Acknowledgements

This chapter is adapted from a completed manuscript coauthored with Joseph F. Ellis,

Amelia E. Triplett, Stanislav S. Rubakhin, and Jonathan V. Sweedler, submitted for publication in 2019. Dr. S.S. Rubakhin and A.E. Triplett aided in sample preparation, while J.F. Ellis aided in data analysis and wrote all scripts related to the manuscript. A.E. Triplett also aided in lipid assignments. Prof. J.V. Sweedler aided in manuscript writing and concept development. The authors gratefully acknowledge support from the National Institutes of Health, Award Number

P30 DA018310 from the National Institute on Drug Abuse, and from the National Institute of

Mental Health, Award Number 1U01 MH109062. E.K. Neumann acknowledges support from the National Science Foundation Graduate Research Fellowship Program and E.K. Neumann and

J.F. Ellis acknowledge the Springborn Fellowship. The authors also thank Dr. Troy J. Comi for useful discussions and Dr. Shannon Cornett for custom Bruker Corporation script to aid in exportation of data.

6.1 Introduction

Single cell measurements are imperative for understanding the function and microstructure of highly complex biological systems. Further, representative single cell analysis within such complex biological systems is further complicated by the sheer number of cells, specifically for the mammalian brain that contains over a trillion cells.[1] These cells are often morphologically, functionally, and/or chemically defined and can be categorized into cell classes and subclasses.[2-

9]

156

A robust classification and sub-classification of individual cells in large cellular populations beyond traditional immunocytochemical and electrophysiological profiles has primarily been spearheaded by the single cell transcriptomics community, and the transcriptomics data has helped to uncover and explore important neurological functions.[10] However, the specific lipids found within a cell depends on a complex interplay between the cellular enzymes encoded by the cell transcriptome,[11] the cellular environment,[12] and organismal metabolic state.[13-15] The ability to probe lipid heterogeneity within the brain has lagged behind the ability to characterize other molecular classes at the single cell level, because of their inherent complexity, and the difficulty creating selective immunohistochemical and fluorescence probes.[15]

Single cell mass spectrometry (MS) methods have been at the forefront of chemical analysis for decades and helped decipher the roles of many different functional molecules, such as neuropeptides.[16] MS approaches compliment transcriptomics efforts by detecting the physical presence of endogenous metabolites, peptides, and proteins rather than the gene expression profiles that suggest cellular potential rather than actual chemical contents. Recently, high- throughput single cell matrix-assisted laser desorption/ionization (MALDI) MS approaches have been developed to characterize chemical heterogeneity of cellular populations by detecting tens to hundreds of chemicals within a single cell. The fast acquisition rate (on the order of seconds) allows for thousands of cells to be profiled in a day, proving the method capable of multiplexed, chemical investigation of complex and abundant biological systems.[17-20]

Typical lipidomics experiments survey the lipid content through sampling larger samples but this does not necessarily detect lipids that are in a small subset of cells within the system;[21] nor provide information on their distributions. Because transcriptomics provides information on gene expression and antibodies are not available to probe specific lipids found within the cellular

157 membrane, there is currently no available approach to obtain such information. LC-MS allows the characterization of the brain lipidome,[22] but provides little information on their distribution, and may miss lipids that are at significant levels within a cell. MS imaging (MSI) is well suited for lipid analysis on a pixel-by-pixel basis, and recent advances have allowed MSI to reach cellular resolutions, although it is difficult to perform these measurements without custom instrumentation.[23,24] In addition, even when MSI obtains a pixel size at the level of (or even better than) single cells, more than one cell is likely interrogated due to the highly intertwined nature of brain cells (e.g. neurons and astrocytes). By using well-established dissociation methods and depositing cells diffusely, we ensure that only individual cells are sampled.[19,20]

Here, we perform microscopy-guided profiling to probe over thirty thousand cells with

MALDI Fourier transform ion cyclotron resonance (FT-ICR) MS to explore the lipid heterogeneity within the cells of the rat cerebellum. To date, this is the largest number of cells profiled within a single cell MS experiment. Additionally, FT-ICR MS measurements provide high-resolution and high-mass accuracy measurements that aid in assignment of over 500 putative lipid spectral features. As tandem MS of the compounds within each cell is problematic given the small amount present, especially when identifying low-abundance, rare chemical species, we assigned our lipid peaks based on liquid chromatography (LC)-MS databases and data repositories of lipids.

6.2 Experimental Section

Chemicals

Chemicals were purchased from Sigma Aldrich and used without further purification unless otherwise specified.

158

Animals

15 2–2.5-month-old male Sprague Dawley outbred rats (Rattus norvegicus)

(www.envigo.com) were housed on a 12-h light cycle and fed ad libitum. Animal euthanasia was performed in accordance with the appropriate institutional animal care guidelines (the

Illinois Institutional Animal Care and Use Committee), and in full compliance with federal guidelines for the humane care and treatment of animals.

Tissue Dissociation and Preparation of Individual Cell Populations

Dissected cerebellum tissues were treated with the papain dissociation system

(Worthington Biochemical, Lakewood, NJ). Briefly, tissues were incubated with an oxygenated solution of 20 units/mL of papain, 1 mM L-cysteine, and 0.5 mM ethylenediaminetetraacetic acid dissolved in Earle's balanced salt solution for 80 min at 34 °C. Tissue was then mechanically dissociated in modified Gey’s balanced salt solution (mGBSS) containing (in mM): 1.5 CaCl2, 5 KCl, 0.2 KH2PO4, 11 MgCl2, 0.3 MgSO4, 138 NaCl, 28 NaHCO3, and 0.8

Na2HPO4, and 25 HEPES, pH 7.2 and supplemented with 0.04% paraformaldehyde (PFA) for cell stabilization through mechanical dissociation. Following dissociation, glycerol was added to a final concentration of 40% (v/v) and 50 µL of cell suspension was transferred onto non-coated microscopy glass slides (LabMed, Ft. Lauderdale, Florida) that had been etched with fiducial marks and washed with ethanol prior to plating.

Each slide held cells from three biological replicates plated in duplicate in a random coordinate on a gridded pattern. This design of experiments helps control batch variation and possible position-dependent artifacts. After ~18 h, cells were stained with Hoechst 33342 (1

µg/mL in mGBSS) for 15 min before quickly rinsing twice with 500 µL of 150 mM ammonium acetate. In total fifteen animals across fifteen slides were analyzed for approximately 35,000

159 cells. Five thousand objects were later removed because they did not have insufficient lipid signal and were assumed to be cellular debris and other fluorescent objects.

Optical Imaging

Brightfield and fluorescence images were acquired on a Zeiss Axio M2 microscope

(Zeiss, Jena, Germany) equipped with an Ab cam Icc5 camera, X-cite Series 120 Q mercury lamp (Lumen Dynamics, Mississauga, Canada), and a HAL 100 halogen illuminator (Zeiss, Jena,

Germany). The DAPI (ex. 335-383 nm; em. 420-470 nm) dichroic was used for fluorescence excitation. The images were acquired with a 10x objective (1-pixel width is 0.55 µm) with a 13% overlap produced during image tiling. Images were processed and exported on ZEN software version 2 blue edition (Zeiss, Jena, Germany). Images were exported as big tiff files.

Sample Preparation Including MALDI Matrix Application and MS Analysis

Slides were coated with a 50 mg/mL DHB solution dissolved in 1:1 ethanol: water with

0.1% trifluoroacetic acid using an automatic sprayer described previously.[18] The matrix solution was nebulized at 10 mL/hr using N2 gas at 50 psi with 100 passes. Samples were taped to a rotating plate and the spray was placed 3 cm above the samples. The total amount of matrix applied was between 0.1 and 0.2 mg/cm2.

Single cell MALDI mass spectra were acquired on the solariX XR 7T FT-ICR mass spectrometer (Bruker Corp., Billerica, MA) with a mass window of 150-3000, yielding a transient length of 2.94 sec. The ultra large laser setting was used with a laser footprint of approximately 100 µm. 50 laser shots were accumulated with a frequency of 1000 Hz at 50% of the maximum power settings. Single cell coordinates and a custom geometry file were ascertained from microMS, as previously described,[20] as well as an Excel file for the target automation function on ftmsControl (v. 2.1.0, Bruker Corp.). Cells were filtered by distance to

160 remove cells that were not 100 µm apart and size to remove cell nuclei. The acquisition was randomized across all three animals on the slide. Further instrumental parameters are supplied in a text document in the SI. Single Cell mass spectra are made publicly available for download at the Illinois Databank.

Cerebellum tissue was a homogenized within a solution of 1:2 chloroform:menthanol

(v:v). Contaminates were removed by centrifuging with water and the organic layer was dried down. Lipid extract was resuspended in a 1:1 ratio of chloroform to a solution of MALDI matrix

(50 mg/mL DHB solution dissolved in 1:1 ethanol: water with 0.1% trifluoroacetic acid) and spotted onto a Bruker 364 spot target.

MALDI-MS of lipid extract was acquired using the same mass window, laser size, and frequency. 150 laser shots were accumulated at 60% of the maximum power settings. Direct infusion electrospray ionization (ESI) of a cerebellum lipid extract was performed for comparison with single cell results. ESI flow rate was 150 µL/hr and a capillary voltage of 3900

V.

Data Analysis and Statistics

Single cell mass spectra were converted into Bruker .d into .xml files using a custom script using the Automation Engine within Data Analysis software (v. 4.4, build 200.55.2969,

Bruker Corp.). Data was then imported into MATLAB 2017A (The MathWorks, Inc., Natick,

MA) using a custom script to parse individual xml files for m/z value, intensity, and resolution.

Data for each cell were combined into a single dataset. Each slide was independently recalibrated with [PC(32:0)+H]+, [PC(34:3)+H]+, and [PC(34:1)+H]+ at m/z 734.56943, 756.55378, or

760.58416, respectively, to correct for mass shifts. Cells not containing at least one of these lipids were excluded for further analysis. Single cell mass spectra were aligned with non-uniform

161 piecemeal bin widths—as mass resolution changes as a function of m/z value for FT-ICR MS.

For example, the bin width at m/z 150 was 0.001 and 0.05 Da at m/z 2500, with 10 additional divisions between. Bin widths were estimated as the average peak width at m/z values and constructed as piecewise linear divisions over the entire spectral range. All intensities were root- mean-square (RMS) normalized.

Peaks between m/z value of 700-900 with mass defects greater than 0.3 Da or less than

0.8 Da were kept for multivariate statistical analysis. The intensity matrix from the selected lipid range was imported into R, where the matrix was mean-centered and z-scored prior to dimensionality reduction using principal component analysis (PCA). 2-D visualization was performed using the first 100 PC scores for t-distributed stochastic neighbor embedding (t-SNE) with a perplexity of 30 and max iterations of 2500.48 Clustering was performed on the t-SNE

Euclidean coordinates using the Louvain clustering method weighting each edge in the graph by the Jaccard overlap index. R scripts for this method were obtained and modified from Shekhar et al. 2016.49,50

To determine which t-SNE clusters were more similar to neurons or astrocytes, we incorporated a neuron/astrocyte relative enrichment spectrum as a training set from the previous chapter (< 3 ppm error). A Pearson pairwise linear correlation was used to compare the reduced spectrum to the relative enrichment training spectrum and the correlation matrix was subsequently mean-centered and plotted with a cutoff similarity of 5%.

Lipid class information was derived by concatenating the assigned lipids into a single matrix containing all lipids within that class and then summing the intensities after normalizing to the maximum lipid intensity within that class. The relative abundance of each respective lipid

162 class was plotted on the t-SNE plot using the “Hot” color map (MATLAB) between the normalized minimum (0) and maximum intensity (64).

Rare cell finding methods were based on Ong et al. Anal. Chem 2015,[17] where lower principal components were used to target unique populations that contributed to lower degrees of variance. In MATLAB, the RMS normalized intensity matrix for the lipid range was mean- centered for PCA. Using only the PCs contributing to the top 90% of the total variance, the PC scores were back-projected and subtracted from the preprocessed spectra resulting in a

‘difference mass spectrum’.

Lipids were subsequently assigned using high mass accuracy, LIPID MAPS structural computational database,[26] and previously published biological data (i.e. lipids found only within bacterial systems were excluded). All m/z values containing less than 10 cells were removed from selection. The 10 m/z values contributing most to the difference mass spectrum were selected as rare lipids and labeled with their putative lipid identity from LIPID MAPS. Lipid assignments all consisted of errors less than 5 ppm, although most assignments consisted of errors less than 3 ppm. Lipids found in less than 10 cells were removed from the study because these spectral features were likely a result of a small mass shift or contamination and, typically, not found between more than one biological replicate.

6.3 Results and Discussion

Single Cell MALDI FT-ICR Analysis

We dissociate our cells onto a slide, locate these features and probe them via MALDI MS using microMS.[20] We first image the slide and determine the locations of the cell nuclei; we filter the fluorescent nuclei to exclude structures located less than 100 µm apart and objects that appear non-cellular by shape and size (e.g. debris, broken cells, and bare nuclei). Filtering by

163 distance and size reduce cell-to-cell contamination of the single cell spectra. We then acquire mass spectra on each of the thousands of remaining cells. An advantage of this approach is that only cell-like features are measured; a typical MSI experiment spends most of its time sampling the space between dispersed cells or may probe only a portion of the cell depending where the cell is within the MSI rastering pattern. Once the spectra are obtained, spectral features with mass defects between 0.3 and 0.8 m/z values were used to find lipid candidates. Each selected m/z value was then matched against the LIPID MAPS structural computational database[25,26] to assign putative identifies for each spectral feature. Only candidates with mass errors lower than 5 ppm were considered, although most errors were below a 3 ppm threshold.

Using this procedure, we detected 520 distinct spectral features within the lipid mass region of ~30k rat cerebellum cells (Table 6.1), with the individual spectra available at the

Illinois databank. The number of detected features is an order of magnitude higher than the number of lipid features detected within a lipid extract analyzed by either MALDI FT-ICR MS

(34 lipid features) or ESI FT-ICR MS (52 lipid features; Figure 6.1). Typical lipidomics experiments survey the lipid content through comprehensive lipid extractions followed by LC-

MS.21 While lipid extracts from larger structures allows one to characterize hundreds of lipids, it does not provide details on cell heterogeneity nor details on lipids found only in the occasional cell, even if the lipid is at high levels in that cell. The single cell approach circumvents these issues, especially for low-abundance compounds found at higher levels within a subpopulation of cells. The largest limitation is limiting identifications to LC-MS based databases because of our inability to perform high quality tandem MS.

On average we detected 34 lipids within each cell (ranging between 1 and 116) from these different lipid classes: phosphatidylcholine (PC), phosphatidylinositol (PI),

164 phosphatidylethanolamine (PE), phosphatidylserine (PS), triglyceride (TG), diglycerides (DG), hexcerimides, and phosphatidic acids (PA), demonstrating the capacity to detect many types of molecules using single cell MALDI MS. In total, 297 of filtered spectral features were matched to known lipids within the LIPID MAPS database with the aforementioned criteria.

As expected for positive-mode MALDI MS, most of the assigned lipids were phospholipids likely originating from the outer plasma membrane. Differences within the abundance of these lipids impact membrane fluidity and curvature; the lipid composition can be correlated to cellular outgrowth and morphology. Other lipids are involved within cell-to-cell signaling or cellular division. The rich diversity of lipids has been noted within literature,[27-29] and our capacity to detect many different types of lipids is necessary for beginning to understand lipid heterogeneity on cellular resolutions. Moreover, the clusters identified within this dataset are primarily a result of natural variations within the cerebellum as opposed to animal-to-animal differences or sample preparation because neither of these experimental variables explain the extent of clustering found within the dataset (Figure 6.1).

Lipid Clustering and Classification

We use the detected lipid features to cluster individual cells using t-distributed stochastic neighbor embedding (t-SNE) with subsequent Louvain-Jaccard clustering. Overall, 101 distinct clusters are identified by the Louvain-Jaccard clustering criteria (Figure 6.2A). There are less than 30 canonical cell types within the cerebellum,[30,31] so obtaining 101 clusters shows the sensitivity of our analytical approach. Cluster 1 was the largest cluster and contained 882 cells, while clusters 100 and 101 were the smallest with 31 cells each.

While our experimental conditions prevent directly correlating our clusters to ICC profiles of canonical cell types within the brain (e.g. neurons and glia), we characterize a

165 surprisingly large lipid diversity and distribution within the cerebellum. To better relate this diversity to brain chemistry, we can use previously ascertained lipid profiles of neurons and astrocytes to determine if a cluster is more similar to neurons, astrocytes, or neither (Figure

6.2B). Even though the lipid profiles were determined using different instrumentation and animals, we found clusters that correlate highly with either neurons or astrocytes, such as cluster

11 and 39. Because ICC was not performed on the cells discussed here, we believe that MS lipid profiles themselves could be sufficient for independent classification of canonical cells within the brain. This is of particular interest, because of the inherent difficulties and limitations of antibody classifications. Moreover, each cluster is chemically different, so the various neuron- like clusters could represent neuronal subclasses that differ by their lipid constituents (e.g. clusters 39 and 72). A similar statement could be said for astrocytes as well as the clusters that do not correlate with either neurons or astrocytes. The clusters that do not correlate with astrocytes nor neurons are likely a mixture of glia (excluding astrocytes), oligodendrocytes, neurons or astrocytes that would not have stained in the previous study, and other cell types.

Some of these clusters could likely be classified with additional MS training from other ICC- based classifications and is certainly a future direction of follow-up studies using this approach.

By recoloring each cluster within the t-SNE plot, we can visualize the distribution of specific lipid classes and use this information to begin to parse out why each of these clusters are different. For example, PA lipids are higher abundance within cluster 78 compared to all other lipid classes and TG lipids are in higher abundance within clusters 99 and 49 compared to DG and PS lipids (Figure 6.3). While the distribution for each of the lipid classes can be ascertained, we only show a graphic for PC (Figure 6.2C), SM (Figure 6.2D), and PG (Figure 6.2E) lipids for clarity; however, the other classes can be found within Figure 6.3. PC lipids appear ubiquitous

166 within the profiled cells, which is in good agreement with their common and relatively high abundance within cell membranes, as well as their high ionization efficiency.[32-34] Perhaps what is even more interesting is the overlap between neuron-like cells and clusters containing a high abundance of SM lipids (e.g. clusters 11, 62, and 69 Figure 6.2D), as SM lipids are well known to compose neurons and are implicated in diseases that affect neurons.[35-38] While certain SM lipids are present in most cellular membranes, the ones localized to neuron-like clusters may be related to specific neuron morphology and function. PG lipids, however, are not as widespread as either PC or SM lipids and are localized to only a few clusters, such as clusters 5, 25, and 68

(Figure 6.2E). Because none of the lipid classes have identical distributions within cells and their subsequent clusters, t-SNE allows us to visualize this heterogeneity within the profiled cells and their relative abundance.

Average Cluster Spectra and Rare Lipid Analysis

Further, we can begin discriminating the lipid profile of each cluster by averaging the spectra acquired from each corresponding cell (Figure 6.4 and 6.5). Of the 520 spectral features detected within the cells, 57% of the compounds were not be matched to known lipids and are the emphasis of follow-up studies. [PC(32:0)+H]+ (m/z = 734.5712) and [PC(34:1)+H]+ (m/z =

760.5845) correspond to spectral features that are present within 99% and 90% of cells, respectively, indicating that these are highly conserved lipids as has previously been shown.39-41

+ + There are others, such as putative [PE(42:8)+H-H2O] and [PG(40:2(OH))+Na] , which are only present within a few clusters because they are only observed in a small number of cells. Each cluster has distinct lipid profiles, some of which vary quite wildly, such as clusters 28 and 91 that clearly differ in number of lipid signatures and their relative intensities (Figure 6.4), especially with [HexCer(t34:2)+K]+ (m/z 752.5074) and [PC(34:1)+Na]+ (m/z 782.5652). While average

167 mass spectra for only eight clusters are displayed in Figure 6.4, similar results are obtained for the other clusters and are illustrated within the Figure 6.5.

Moreover, we can use our rich dataset to find lipids that are uncommon and detected within a small number of cells. Using the method described in detail within the methods section, we determined 10 spectral features that detected in less than 6.5% of cells with one unknown compound present in as little as 0.1% of the cells (Figure 6.6A). Because of their relatively low abundance, we assume that these lipids are often overlooked using traditional visualization

+ methods. A peak for putative [PE(42:8)+H-H2O] at an m/z value of 798.5414 is shown along with its C13 peak (m/z 799.5453), demonstrating that this is a rare lipid rather than an artifact

(Figure 6.6B top) and for putative [PG(40:2(OH))+Na]+ at an m/z value of 869.5807, with the exception that its C13 peak is not completely resolved from the peak at m/z 870.5856 (Figure

6.6B bottom). Nonetheless, this single cell MALDI MS method can be used to not only find rare lipid features among hundreds of more highly abundant chemical species but also locate these interesting cells for follow-up analysis with other analytical approaches.[42-44]

We have demonstrated the utility of high throughput single cell MALDI MS for in depth chemical characterization of single mammalian cells. In sum, we detect 520 lipid features within thirty thousand individual cells using MALDI FT-ICR MS. From the information gained here, we are able to determine 101 significantly different clusters ranging in size using a combination of t-SNE and Louvain-Jaccard clustering, correlate some of the resulting clusters to previously ascertained astrocytic and neuronal lipid markers and visualize the localization of different lipid classes. Each cluster is represented by different lipid profiles, while still having some conserved lipid features, such as PC(32:0). Some of the spectral features were located only in a small number of cells, perhaps relating to rare functions possessed by brain cells.

168

We focused on lipid profiling within this study due to the functional importance of lipids within the brain and as well as the significant technical difficulties surrounding lipidomics, particularly at cellular resolutions.[45-47] Our ability to characterize thousands of cells for their lipid contents will enable lipid heterogeneity to be studied at an unprecedented level, allowing the interactions between cellular phenotype to be linked to the transcriptome and organismal state. Of course, the approaches can be expanded to include additional molecular classes such as metabolites, peptides, and proteins. A last exciting ability is to target the rare cells highlighted here for follow-up analysis with other analytical approaches such as CE-MS[42] and vibrational spectroscopy,[44] as the MALDI measurement leaves enough material behind to enable such measurements and multimodal analysis has demonstrated enhanced classification schemes.[43]

While the described approach can easily be optimized for other brain regions and other animal models, the future of these experiments lies in the quality and abundance of chemical information held within each single cell that is acquired and can be used to tackle the poorly understood lipidome. Using this information, we can relate lipid profiles to known cellular functional and morphological differences among canonical cell types that has yet to be understood or even studied.

6.4 Conclusions

We have demonstrated a method for high-throughput analysis of the 520 lipids within thirty thousand individual cells using MALDI FT-ICR MS. From the information gained here, we are able to determine 101 significantly different clusters using a combination of T-SNE and

Louvain-Jaccard clustering ranging in size. Each cluster is markedly different, while still having some conserved lipid features, such as PC(32:0). Some of the spectral features were located only in a small number of cells, perhaps relating to rare functions possessed by brain cells. Ultimately,

169 we have developed a protocol that allows for chemical exploration of the chemical heterogeneity within the brain, although we focus on the lipidome here. We focused on lipids because they have been known to have vast functional importance within the brain, but truly little is known about them.[24-26] While this method can easily be optimized for other brain regions and other animal models, the future of these experiments lies in the quality and abundance of chemical information held within each single cell that is acquired and can be used to tackle the poorly understood lipidome. Using this information, we can start ascertaining the functional relationship to this chemical heterogeneity and relate it to functional and shape differences among canonical cell types that has yet to be understood or even studied. Further work will involve identifying the structure and function of these lipids in addition to testing if these spectral features are more robust classification criteria than traditional immunohistochemical markers, such as glial fibrillary acidic protein or neurofilament L.

6.5 Acknowledgements

The authors gratefully acknowledge support from the National Institutes of Health,

Award Number 1U01 MH109062 from the National Institute of Mental Health and Award

Number R01 AI113219 from the National Institute of Allergy and Infectious Diseases, and P30

DA018310 from the National Institute on Drug Abuse. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of

Health. E.K.N. acknowledges support from the National Science Foundation Graduate

Research Fellowship Program and E.K.N and J.F.E acknowledge support from the Springborn

Fellowship. The authors would also like to thank Dr. Shannon Cornett for developing the

170 custom Bruker DataAnalysis script for data exportation and Dr. Troy Comi for useful discussions.

171

6.6 Figures and Table

Figure 6.1: A) Electrospray ionization MS and B) MALDI MS lipid extract average spectra.

172

Figure 6.2 A) Single cells within the rodent cerebellum can be classified into 101 clusters based on their spectral features within the lipid mass range. The clustering is based on lipid content and/or relative signal intensity. Information from ~30k single cells are visualized using a t-SNE plot and clustered using the Louvain-Jaccard method. Each cluster is colored and numbered.

Cluster 1 contains the most cells (882 cells), while clusters 100 and 101 contain the least (31 cells) in each. B) t-SNE plot recolored to show correlation of individual cells to neuronal and astrocytic lipid profiles that have been previously determined. C) t-SNE recolored by relative amount of phosphatidylcholine, D) sphingomyelin and E) phosphatidylglycerol lipid species to visualize the localizations of lipid classes.

173

Figure 6.3 t-SNE plot of cells colored by analysis date. FT-ICR MS analysis was performed over 22 days, where each day is represented by an individual color.

174

Figure 6.4 Average spectra of all cells located within their respective clusters. Visual observation of the clusters shows detectable spectral diversity that can be extended to all 101 clusters. All spectral assignments are located in in Table 6.1 in m/z order. Only eight are shown here for clarity.

175

Figure 6.5 (cont.)

176

Figure 6.5 (cont.)

177

Figure 6.5 (cont.)

178

Figure 6.5 (cont.)

179

Figure 6.5 cont.)

180

Figure 6.5 (cont.)

181

Figure 6.5 (cont.)

182

Figure 6.5 (cont.)

183

Figure 6.5 (cont.)

184

Figure 6.5 (cont.)

185

Figure 6.5 (cont.)

186

Figure 6.5 (cont.)

187

Figure 6.5 (cont.)

188

Figure 6.5 (cont.)

189

Figure 6.5 (cont.)

190

Figure 6.5 (cont.)

191

Figure 6.5 (cont.)

192

Figure 6.5 (cont.) Average mass spectra for each of the 101 clusters with the highest root-mean square normalized intensities labelled with their respective m/z value.

193

Figure 6.6 A) Spectra features that are detected within less than 6.5% of the analyzed cerebral cells as determined by reverse principal component analysis. A variety of lipid classes are represented within the plot. Some features are detected in less than 0.1% of analyzed cells. B)

Two single cell mass spectra that highlight two of the rare spectral features, demonstrating that the features have adequate signal-to-noise ratios and expected isotopic patterns. The insets indicate that these are real spectral features rather than random noise or spectral artifacts.

194

Molecular Error Number Measured m/z Exact m/z Assignment Formula (ppm) of Cells 700.3737 not matched 14 + 700.4568 700.454796 C37H67NO9P [PE(32:4(OH))+H] 2.8610 3794 + 700.4752 700.475925 C34H71NO11P [PG(28:0(OH))+NH4] -1.0350 13 700.5093 not matched 172 + 701.4621 701.462310 C41H65O9 [MGDG(32:6)+H-H2O] -0.2994 79 + 701.4848 701.486430 C37H70N2O8P [PE(32:4)+NH4] -2.3236 1095 702.4883 not matched 246 + 703.5006 703.502080 C37H72N2O8P [PE(32:3)+NH4] -2.1038 206 703.5811 not matched 115 + 704.5838 704.582366 C43H78NO6 [TG(40:4)+NH4] 2.0352 16 705.3712 not matched 159 705.5234 not matched 3152 + 706.3747 706.377333 C30H61NO15P [MIPC(d18:0)+H] -3.7275 31 706.5077 not matched 20151 706.5269 not matched 519 706.5418 not matched 4776 + [LPS(30:2)+NH4] + 707.4946 707.496995 C36H72N2O9P [PC(28:2(OH))+NH4] -3.3852 18 + [PS(P-30:1)+NH4] + 707.5086 707.509260 C41H71O9 [MGDG(32:3)+H-H2O] -0.9328 14423 707.5384 not matched 2736 707.5462 not matched 12 + 707.6687 707.670058 C49H87O2 [CE(22:1)+H] -1.9190 15 708.4859 not matched 26 [PE(32:0(OH)) +H]+ 708.5139 708.517396 C37H75NO9P + -4.9342 53 [PA(34:1(OH))+NH4] 708.5419 not matched 325 709.4011 not matched 334 [PA(O-38:5)+H]+ [PA(P-38:4)+H]+ + [PA(38:3)+H-H2O] 709.5166 709.516668 C41H74O7P [PA(O-38:4(OH))+H- -0.0958 17 + H2O] [PA(P-38:3(OH))+H- + H2O] 710.4046 not matched 40 710.4650 not matched 5735 710.5008 not matched 775 711.4668 not matched 922 + 711.5044 711.504760 C37H73N2O7PNa [SM(t32:2)+Na] -0.5060 25 712.4817 not matched 74

Table 6.1 (cont.)

195

Molecular Error Number Measured m/z Exact m/z Assignment Formula (ppm) of Cells + 713.3855 713.387170 C33H62O14P [PI(24:1(OH))+H] -2.3409 113 [PA(34:1)+K]+ + 713.4546 713.451814 C37H71O8PK [PA(O-34:2(OH))+K] 3.9050 11 [PA(P-34:1(OH))+K]+ + 713.5990 713.595587 C41H82N2O5P [SM(d36:1)+H-H2O] 4.7828 10 714.3689 not matched 142 + 714.3794 714.382419 C32H61NO14P [MIPC(d20:1)+H-H2O] -4.2260 518 715.3872 not matched 843 + 715.4004 715.402820 C33H64O14P [PI(24:0(OH))+H] -3.3827 35 + 715.4179 715.418076 C37H64O11P [LPI(28:4)+H-H2O] -0.2460 348 + 716.3900 716.389942 C34H64NO10PK [PS(28:1)+K] 0.0810 642 716.4512 not matched 11118 + 716.4608 716.462561 C39H68NO7PNa [LPE(34:6)+Na] -2.4579 27 717.3830 not matched 45 717.3935 not matched 88 + 717.4539 717.454819 C39H66O10Na [MGDG(30:4)+Na] -1.2809 1121 [PS(30:3(OH))+H]+ 718.4286 718.428975 C36H65NO11P + -0.5220 42 [PG(30:5(OH))+NH4] + 719.5003 719.496995 C37H72N2O9P [PE(32:3(OH))+NH4] 4.5935 39 [PA(O-40:5(OH))+H- H O]+ 719.5371 719.537403 C H O P 2 -0.4211 21 43 76 6 [PA(P-40:4(OH))+H- + H2O] 720.5581 not matched 26 + 721.5092 721.512645 C37H74N2O9P [PE(32:2(OH))+NH4] -4.7747 29 721.6868 not matched 36 + 722.3649 722.364121 C32H62NO12PK [PI-Cer(t26:1)+K] 1.0784 45 722.4996 not matched 10499 722.5206 not matched 12 722.6895 not matched 23 [PA(38:5)+H]+ + 723.4953 723.495932 C41H72O8P [PA(O-38:6(OH))+H] -0.8735 6000 [PA(P-38:5(OH))+H]+ + 723.5023 723.501769 C39H72O10Na [MGDG(30:1)+Na] 0.7339 1258 + 724.4787 724.476041 C38H71NO9K [HexCer(t32:2)+K] 3.6702 19754 724.4989 not matched 1315 724.5155 not matched 411 725.4812 not matched 12585 + 725.5112 725.511582 C41H74O8P [PA(38:4)+H] -0.5265 2137

Table 6.1 (cont.)

196

Molecular Error Number Measured m/z Exact m/z Assignment Formula (ppm) of Cells + 725.5587 725.556796 C39H79N2O6PNa [SM(d34:1)+Na] 2.6242 20 [CerP(t38:2)+K]+ + 726.4849 726.483448 C38H74NO7PK [LPC(30:2)+K] 1.9987 763 [PC(P-30:1)+K]+ 726.5137 not matched 65 + 727.4623 727.465695 C38H68N2O9P [LPS(32:6)+NH4] -4.6669 132 + 727.5011 727.502080 C39H72N2O8P [PE(34:5)+NH4] -1.3471 102 727.5861 not matched 13 + 728.5048 728.507225 C36H75NO11P [PG(30:0(OH))+NH4] -3.3287 12 + [PE(34:4)+NH4] 729.5172 729.517730 C39H74N2O8P + -0.7265 38 [PE(P-34:4(OH))+NH4] + [PE(34:3)+NH4] [PE(O- 731.5346 731.533380 C39H76N2O8P + 1.6677 88 34:4(OH))+NH4] + [PE(P-34:3(OH))+NH4] [SM(d36:1)+H]+ 731.6084 731.606151 C41H84N2O6P + 3.0740 25552 [SM(t36:0)+H-H2O] 732.3545 not matched 1113 732.4433 not matched 2177 [PS(32:2)+H]+ [PI-Cer(d32:2)+H- H O]+ 732.4808 732.481011 C H NO P 2 -0.2881 57 38 71 10 [PS(32:1(OH))+H- + H2O] + [PG(32:4)+NH4] + 732.5558 732.556151 C47H74NO5 [DG(44:11)+NH4] -0.4791 5859 + 732.6108 732.613666 C45H82NO6 [TG(42:4)+NH4] -3.9120 20487 + 733.3583 733.361610 C39H57O11S [SQDG(30:8)+H-H2O] -4.5135 674 + 733.4470 733.443897 C41H66O9P [PA(38:8(OH))+H] 4.2307 62 + [PE(34:2)+NH4] [PE(O- 733.5470 733.549031 C39H78N2O8P + -2.7687 370 34:3(OH))+NH4] + [PE(P-34:2(OH))+NH4] 733.5582 not matched 606 + 733.6145 733.612937 C49H81O4 [DG(46:7)+H-H2O] 2.1306 1687 734.3569 not matched 39 734.3657 not matched 16 [PE(34:1(OH))+H]+ 734.5344 734.533046 C39H77NO9P + 1.8433 2011 [PA(36:2(OH))+NH4] + 734.5707 734.569432 C40H81NO8P [PC(32:0)+H] 1.7262 29772 + 735.3781 735.377260 C39H59O11S [SQDG(30:7)+H-H2O] 1.1423 48

Table 6.1 (cont.)

197

Molecular Error Number Measured m/z Exact m/z Assignment Formula (ppm) of Cells + 735.5381 735.540560 C43H75O9 [MGDG(34:3)+H-H2O] -3.3445 17 735.5744 not matched 29185 736.5781 not matched 24055 + 737.5030 737.499433 C39H75N2O6PK [SM(d34:3)+K] 4.8366 41 + [PE(34:0)+NH4] [PE(O- 737.5805 737.580331 C39H82N2O8P + 0.2291 803 34:1(OH))+NH4] + [PE(P-34:0(OH))+NH4] 738.4305 not matched 27 738.4930 not matched 4816 + 738.5142 738.515074 C41H72NO10 [MGDG(32:5)+NH4] -1.1834 27 [PA(36:2)+K]+ + 739.4679 739.467464 C39H73O8PK [PA(O-36:3(OH))+K] 0.5896 12 [PA(P-36:2(OH))+K]+ 739.4954 not matched 790 + 740.4716 740.470840 C36H71NO12P [PI-[Cer(t30:1)+H] 1.0264 7351 + 741.4316 741.433726 C39H66O11P [LPI(30:5)+H-H2O] -2.8674 416 741.4753 not matched 476 743.4928 not matched 60 743.5790 not matched 11 744.4090 not matched 15 744.4790 not matched 106 744.5715 not matched 15 [LPG(32:1)+K]+ + 745.4789 745.478029 C38H75O9PK [PG(O-32:1)+K] 1.1684 145 [PG(P-32:0)+K]+ 745.5077 not matched 1082 [PI-[Cer(d30:1)+Na]+ 746.4589 746.457869 C H NO PNa 1.3812 13116 36 70 11 [PS(30:0(OH))+Na]+ 746.4951 not matched 66 + 746.5114 746.511917 C43H73NO7P [PE(38:6)+H-H2O] -0.6925 275 [CerP(t42:1)+H]+ [PC(O-34:1)+H]+ + 746.6064 746.605817 C42H85NO7P [PC(P-34:0)+H] 0.7809 4334 [PC(O-34:0(OH))+H- + H2O] 747.3689 not matched 51 + 747.4626 747.459547 C42H68O9P [PG(36:7)+H-H2O] 4.0845 4848 + 747.5238 747.528295 C39H76N2O9P [PE(34:3(OH))+NH4] -6.0131 192 747.6101 not matched 1648

Table 6.1 (cont.)

198

Molecular Error Number Measured m/z Exact m/z Assignment Formula (ppm) of Cells [PC(32:1(OH))+H]+ [PS(O-34:1)+H]+ [PS(P-34:0)+H]+ 748.5501 748.548696 C40H79NO9P [PS(O-34:0(OH))+H- 1.8756 16 + H2O] + [PG(O-34:3)+NH4] + [PG(P-34:2)+NH4] [PE(36:0)+H]+ [PE(O-36:1(OH))+H]+ [PE(P-36:0(OH))+H]+ [PA(38:1)+NH ]+ 748.5863 748.585082 C H NO P 4 1.6271 3349 41 83 8 [PA(O- + 38:2(OH))+NH4] [PA(P- + 38:1(OH))+NH4] 748.6126 not matched 25 + 749.3913 749.390657 C35H66O12SK [SQDG(26:0)+K] 0.8580 19 + 749.5100 749.509177 C41H75O8PNa [PA(38:3)+Na] 1.0980 40 749.5888 not matched 147 749.7138 not matched 43 750.5275 not matched 72 [LPS(34:0)+H]+ [PS(O-34:0)+H]+ 750.5663 750.564346 C40H81NO9P + 2.6034 24 [PG(O-34:2)+NH4] + [PG(P-34:1)+NH4] + 750.5850 750.585439 C42H81NO8Na [HexCer(d36:1)+Na] -0.5849 172 750.7175 not matched 30 [PA(38:2)+Na]+ + 751.5249 751.524827 C41H77O8PNa [PA(O-38:3(OH))+Na] 0.0971 591 [PA(P-38:2(OH))+Na]+ 751.5887 not matched 18 751.7212 not matched 10 + 752.5074 752.507341 C40H75NO9K [HexCer(t34:2)+K] 0.0784 18212 752.5287 not matched 146 752.5437 not matched 76 753.4249 not matched 213 + 753.4449 753.446729 C39H71O9PK [PA(36:3(OH))+K] -2.4275 22 753.5112 not matched 10491 + 753.5887 753.588096 C41H83N2O6PNa [SM(d36:1)+Na] 0.8015 15657 754.4286 not matched 20

Table 6.1 (cont.)

199

Molecular Error Number Measured m/z Exact m/z Assignment Formula (ppm) of Cells + 754.5136 754.513360 C38H76NO11S [SHexCer(d32:0)+H] 0.3181 336 [PI-Cer(d32:0)+H]+ 754.5224 754.522991 C40H77NO9K + -0.7833 109 [PG(32:1(OH))+NH4] [PC(32:1)+Na]+ 754.5361 754.535726 C H NO PNa 0.4957 129 40 78 8 [PC(P-32:1(OH))+Na]+ + [LPG(34:0)+NH4] 754.5924 754.595646 C40H85NO9P + -4.3016 5390 [PG(O-34:0)+NH4] 755.3611 not matched 89 + [PE(36:5)+NH4] 755.5336 755.533380 C41H76N2O8P + 0.2912 36 [PE(P-36:5(OH))+NH4] + 755.6661 755.667653 C51H88O2Na [CE(24:2)+Na] -2.0551 23

[PC(32:0)+Na]+ + 756.5524 756.551376 C40H80NO8PNa [PC(O-32:1(OH))+Na] 1.3535 27901 [PC(P-32:0(OH))+Na]+

756.6698 not matched 10 757.5548 not matched 24669 + 757.5636 757.562033 C40H83N2O6PK [PE-Cer(d38:0)+K] 2.0685 63 758.5586 not matched 5831 [PC(34:2)+H]+ + 758.5698 758.569432 C42H81NO8P [PS(O-36:1)+H-H2O] 0.4851 6098 + [PS(P-36:0)+H-H2O] + [PE(36:3)+NH4] [PE(O- 759.5635 759.564681 C41H80N2O8P + -1.5548 35 36:4(OH))+NH4] + [PE(P-36:3(OH))+NH4] 759.5735 not matched 456 [SM(d38:1)+H]+ 759.6385 759.637452 C43H88N2O6P + 1.3796 7344 [SM(t38:0)+H-H2O] + 760.4723 760.467798 C41H72NO7PK [PE(P-36:5)+K] 5.9200 1111 760.4848 not matched 41 760.5773 not matched 2765 [PC(34:1)+H]+ [PC(O-34:2(OH))+H]+ [PC(P-34:1(OH))+H]+ 760.5860 760.585082 C H NO P 1.2070 27947 42 83 8 [PC(34:0(OH))+H- + H2O] + [PS(O-36:0)+H-H2O] 760.6398 not matched 363

Table 6.1 (cont.)

200

Molecular Error Number Measured m/z Exact m/z Assignment Formula (ppm) of Cells [PA(38:5)+K]+ + 761.4485 761.451814 C41H71O8PK [PA(O-38:6(OH))+K] -4.3522 17 [PA(P-38:5(OH))+K]+ + 761.4760 761.475197 C43H70O9P [PA(40:8(OH))+H] 1.0545 69 761.5885 not matched 25796 [PI-Cer(t30:1)+K]+ 762.4323 762.431807 C H NO PK 0.6466 35 36 70 11 [PS(30:0(OH))+K]+ + 762.4535 762.452784 C36H70NO12PNa [PI-Cer(t30:1)+Na] 0.9391 613 [PE(36:1(OH))+H]+ 762.5647 762.564346 C41H81NO9P + 0.4642 32 [PA(38:2(OH))+NH4] 762.5922 not matched 13216 [PC(34:0)+H]+ + 762.6010 762.600732 C42H85NO8P [PC(O-34:1(OH))+H] 0.3514 27529 [PC(P-34:0(OH))+H]+ + [PE(36:1)+NH4] [PE(O- 763.5960 763.595981 C41H84N2O8P + 0.0249 811 36:2(OH))+NH4] + [PE(P-36:1(OH))+NH4] 763.6047 not matched 24782 764.6060 not matched 11193 765.4260 not matched 1923 + 765.5335 765.530733 C41H79N2O6PK [SM(d36:3)+K] 3.6145 19 766.3535 not matched 12 + 766.4284 766.426569 C38H66NO11PNa [PS(32:4(OH))+Na] 2.3890 672 766.5210 not matched 50 768.3734 not matched 485 + 768.5009 768.502140 C38H75NO12P [PI-Cer(t32:1)+H] -1.6135 4258 [CerP(t42:1)+Na]+ + 768.5859 768.587761 C42H84NO7PNa [PC(O-34:1)+Na] -2.4213 999 [PC(P-34:0)+Na]+ + 769.3797 769.384128 C41H63O9PK [PA(38:9(OH))+K] -5.7553 20 + 769.4609 769.465026 C41H70O11P [LPI(32:5)+H-H2O] -5.3622 489 [PG(36:5)+H]+ [PG(P-36:5(OH))+H]+ 769.5047 769.501412 C H O P 4.2729 264 42 74 10 [PG(36:4(OH))+H- + H2O] + 769.5622 769.562033 C41H83N2O6PK [SM(d36:1)+K] 0.2170 18 769.5897 not matched 286 + 770.5659 770.567026 C41H82NO8PNa [PE(36:0)+Na] -1.4613 11 771.4872 not matched 116

Table 6.1 (cont.)

201

Molecular Error Number Measured m/z Exact m/z Assignment Formula (ppm) of Cells 772.4422 not matched 5036 [PC(32:0) + K]+ + 772.5259 772.525313 C40H80NO8PK [PC(O-32:1(OH))+K] 0.7598 3867 [PC(P-32:0(OH))+K]+ [CerP(t44:2)+H]+ [PC(O-36:2)+H]+ [PC(P-36:1)+H]+ [PC(36:0)+H-H O]+ 772.6197 772.621467 C H NO P 2 -2.2870 831 44 87 7 [PC(O-36:1(OH))+H- + H2O] [PC(P-36:0(OH))+H- + H2O] 773.4446 not matched 390 [PA(42:8)+H]+ 773.5097 773.511582 C45H74O8P [PA(42:7(OH))+H- -2.4331 163 + H2O] [PG(34:0)+Na]+ + 773.5297 773.530306 C40H79O10PNa [PG(O-34:1(OH))+Na] -0.7834 128 [PG(P-34:0(OH))+Na]+ 773.6234 not matched 154 774.4559 not matched 2926 [PI-Cer(d32:1)+Na]+ 774.4872 774.489170 C H NO PNa -2.5436 11166 38 74 11 [PS(32:0(OH))+Na]+ 774.5234 not matched 20 774.5322 not matched 10 [PC(34:2(OH))+H]+ [PS(O-36:2)+H]+ [PS(P-36:1)+H]+ + [PS(36:0)+H-H2O] [PS(O-36:1(OH))+H- 774.5622 774.564346 C42H81NO9P + -2.7706 196 H2O] [PS(P-36:0(OH))+H- + H2O] + [PG(O-36:4)+NH4] + [PG(P-36:3)+NH4] [PE(38:1)+H]+ + [PA(40:2)+NH4] [PA(O- 774.6009 774.600732 C43H85NO8P + 0.2169 1315 40:3(OH))+NH4] [PA(P- + 40:2(OH))+NH4] Table 6.1 (cont.)

202

Molecular Error Number Measured m/z Exact m/z Assignment Formula (ppm) of Cells [CerP(t44:1)+H]+ [PC(O-36:1)+H]+ + 774.6347 774.637117 C44H89NO7P [PC(P-36:0)+H] -3.1202 615 [PC(O-36:0(OH))+H- + H2O] + 775.4308 775.431079 C41H69O9PK [PA(38:6(OH))+K] -0.3598 19 775.4589 not matched 18 + 775.4902 775.490847 C44H72O9P [PG(38:7)+H-H2O] -0.8343 3774 775.5699 not matched 16 775.6042 not matched 189 + 775.6386 775.636353 C53H84O2Na [CE(26:6)+Na] 2.8970 81 [PE(38:8(OH))+H]+ 776.4870 776.486096 C43H71NO9P + 1.1642 17 [PA(40:9(OH))+NH4] + 776.5027 776.504820 C38H76NO11PNa [PI-Cer(d32:0)+Na] -2.7302 24 [PC(34:1(OH))+H]+ [PS(O-36:1)+H]+ [PS(P-36:0)+H]+ 776.5777 776.579996 C42H83NO9P [PS(O-36:0(OH))+H- -2.9566 1910 + H2O] + [PG(O-36:3)+NH4] + [PG(P-36:2)+NH4] 777.3824 not matched 35 777.5824 not matched 96 [SM(t38:0)+H]+ + 777.6449 777.648016 C43H90N2O7P [PE(O-38:1)+NH4] -4.0069 46 + [PE(P-38:0)+NH4] 777.7433 not matched 45 + 778.6152 778.616739 C44H85NO8Na [HexCer(d38:1)+Na] -1.9766 13 778.6496 not matched 16 778.7465 not matched 31 + 779.4012 779.404864 C43H65O8PK [PA(40:10)+K] -4.7010 69 779.5480 14 [PI-Cer(d34:1)+H]+ [PS(34:0(OH))+H]+ 780.5371 780.538525 C40H79NO11P + -1.8257 3186 [PI-Cer(t34:0)+H-H2O] + [PG(34:2(OH))+NH4] + 780.5496 780.551376 C42H80NO8PNa [PC(34:2)+Na] -2.2753 187 781.5293 not matched 40 [PE(38:6)+NH4]+ 781.5480 781.549031 C43H78N2O8P [PE(P- -1.3192 909 38:6(OH))+NH4]+

Table 6.1 (cont.)

203

Molecular Error Number Measured m/z Exact m/z Assignment Formula (ppm) of Cells + 781.6199 781.619396 C43H87N2O6PNa [SM(d38:1)+Na] 0.6448 597 [PE(38:5(OH))+H]+ 782.5324 782.533046 C43H77NO9P + -0.8255 26 [PA(40:6(OH))+NH4] [PI-Cer(d34:0)+H]+ 782.5512 782.554291 C42H81NO9K + -3.9499 31 [PG(34:1(OH))+NH4] [PC(34:1)+Na]+ + 782.5652 782.567026 C42H82NO8PNa [PC(O-34:2(OH))+Na] -2.3333 22130 [PC(P-34:1(OH))+Na]+ [HexCer(d40:2)+H]+ 782.6481 782.650445 C46H88NO8 [HexCer(t40:1)+H- -2.9962 30 + H2O] 783.5715 not matched 17724 783.6965 not matched 25 + [PI(O-30:1)+NH4] 784.5356 784.533440 C39H79NO12P + 2.7532 21 [PI(P-30:0)+NH4] + 784.5715 784.569941 C42H83NO9K [HexCer(t36:0)+K] 1.9871 2035 [PC(34:0)+Na]+ + 784.5825 784.582676 C42H84NO8PNa [PC(O-34:1(OH))+Na] -0.2243 19998 [PC(P-34:0(OH))+Na]+ 784.6997 not matched 15 785.5044 not matched 50 785.5872 not matched 9614 786.4560 not matched 2195 786.5903 not matched 260 [PC(36:2)+H]+ [PC(O-36:3(OH))+H]+ + 786.6013 786.600732 C44H85NO8P [PC(P-36:2(OH))+H] 0.7221 11993 + PS(O-38:1)+H-H2O] + PS(P-38:0) +H-H2O] 787.4591 not matched 256 787.5185 not matched 38 787.6013 not matched 4387 [SM(d40:1)+H]+ 787.6654 787.668752 C45H92N2O6P + -4.2556 88 [SM(t40:0)+H-H2O] [PC(36:1)+H]+ [PC(O-36:2(OH))+H]+ [PC(P-36:1(OH))+H]+ 788.6138 788.616382 C H NO P -3.2741 23628 44 87 8 [PC(36:0(OH))+H- + H2O] + [PS(O-38:0)+H-H2O] 789.5342 not matched 826

Table 6.1 (cont.) 204

Molecular Error Number Measured m/z Exact m/z Assignment Formula (ppm) of Cells 789.6201 not matched 20068 [PI-Cer(d32:1)+K]+ 790.4607 790.463107 C H NO PK -3.0451 25 38 74 11 [PS(32:0(OH))+K]+ + 790.4811 790.478363 C42H74NO8PK [PC(34:5)+K] 3.4625 508 [PE(40:7)+H]+ [PE(40:6(OH))+H- 790.5373 790.538131 C45H77NO8P + -1.0512 323 H2O] + [PA(42:8)+NH4] 790.6201 not matched 5319 791.4420 not matched 13 791.5498 not matched 76 [HexCer(d42:2)+H- 792.6686 792.671181 C48H90NO7 + -3.2561 498 H2O] 793.6718 not matched 114 [PS(O-36:0(OH))+H]+ + [PG(36:1)+NH4] [PG(O- 794.5874 794.590561 C42H85NO10P + -3.9781 748 36:2(OH))+NH4] [PG(P- + 36:1(OH))+NH4] [CerP(t44:2)+Na]+ + 794.5999 794.603411 C44H86NO7PNa [PC(O-36:2)+Na] -4.4185 96 [PC(P-36:1)+Na]+ 795.4578 not matched 12 + 795.5422 795.539909 C40H79N2O11S [SHexCer(d34:2)+NH4] 2.8798 12 + [PG(O-40:4)+H-H2O] + 795.5906 795.589833 C46H84O8P [PG(P-40:3)+H-H2O] 0.9641 24 [DG(48:10)+Na]+ 796.4609 not matched 88 + 796.5219 796.523924 C40H78NO12S [SHexCer(t34:1)+H] -2.5410 42 [PC(36:5(OH))+H]+ [PS(O-38:5)+H]+ [PS(P-38:4)+H]+ + 796.5516 796.548696 C44H79NO9P [PS(38:3)+H-H2O] 3.6457 10 [PS(P-38:3(OH))+H- + H2O] + [PG(P-38:6)+NH4] [CerP(t44:1)+Na]+ [PC(O-36:1)+Na]+ 796.6172 796.619061 C H NO PNa -2.3361 37 44 88 7 [PC(P-36:0)+Na]+ [PE(P-39:0)+Na]+ 797.4516 not matched 90

Table 6.1 (cont.)

205

Molecular Error Number Measured m/z Exact m/z Assignment Formula (ppm) of Cells + 798.5423 798.543217 C47H77NO7P [PE(42:8)+H-H2O] -1.1483 1116 799.3642 not matched 20 [PG(36:1)+Na]+ + 799.5454 799.545956 C42H81O10PNa [PG(O-36:2(OH))+Na] -0.6954 137 [PG(P-36:1(OH))+Na]+ + [PE(O-40:4)+NH4] 799.6314 799.632366 C45H88N2O7P + -1.2081 11 [PE(P-40:3)+NH4] 800.5392 not matched 13 + 800.5580 800.558867 C47H79NO7P [PE(42:7)+H-H2O] -1.0829 32 801.3689 not matched 35 [PG(36:0)+Na]+ + 801.5611 801.561606 C42H83O10PNa [PG(O-36:1(OH))+Na] -0.6313 15 [PG(P-36:0(OH))+Na]+ 802.4518 not matched 6386 802.4877 not matched 1872 + 802.5174 802.514749 C44H78NO7PK [PC(P-36:5)+K] 3.3034 193 [PC(36:2(OH))+H]+ [PS(O-38:2)+H]+ [PS(P-38:1)+H]+ + [PS(38:0)+H-H2O] [PS(O-38:1(OH))+H- 802.5987 802.595646 C44H85NO9P + 3.8051 24 H2O] [PS(P-38:0(OH))+H- + H2O] + [PG(O-38:4)+NH4] + [PG(P-38:3)+NH4] 803.4550 not matched 1900 + 803.5300 803.533380 C45H76N2O8P [PE(40:9)+NH4] -4.2064 82 804.3535 not matched 1498 + 804.5113 804.514990 C43H76NO9PNa [PE(38:5(OH))+Na] 25 [PC(38:7)+H]+ [PC(38:6(OH))+H- + 804.5535 804.553782 C46H79NO8P H2O] -0.3505 1359 + [PS(O-40:6)+H-H2O] + [PS(P-40:5)+H-H2O] [PC(36:1(OH))+H]+ [PS(O-38:1)+H]+ + 804.6082 804.611296 C44H87NO9P [PS(P-38:0)+H] -3.8478 87 [PG(O-38:3)+H]+ [PG(P-38:2)+H]+ 805.3566 not matched 726 805.5051 not matched 22

Table 6.1 (cont.)

206

Molecular Error Number Measured m/z Exact m/z Assignment Formula (ppm) of Cells + [PI(O-34:1)+H-H2O] 805.5567 805.558926 C43H82O11P + -2.7633 539 [PI(P-34:0)+H-H2O] [SM(t40:0)+H]+ + 805.6770 805.679316 C45H94N2O7P [PE(O-40:1)+NH4] -2.8746 48 + [PE(P-40:0)+NH4] 805.7739 not matched 16 806.3692 not matched 255 [PI-Cer(d36:2)+H]+ [PS(36:1(OH))+H]+ 806.5520 806.554175 C42H81NO11P + -2.6967 135 [PI-Cer(t36:1)+H-H2O] + [PG(36:3(OH))+NH4] [PC(38:6)+H]+ [PC(P-38:6(OH))+H]+ 806.5708 806.569432 C46H81NO8P + 1.6961 4340 [PS(O-40:5)+H-H2O] + [PS(P-40:4)+H-H2O] + [PC(O-40:4)+H-H2O] 806.6442 806.642203 C48H89NO6P + 2.4757 28 [PC(P-40:3)+H-H2O] 806.6801 not matched 23 807.3739 not matched 40 + 807.5739 807.574577 C43H84O11P [PI(O-34:0)+H-H2O] -0.8383 1492 808.3693 not matched 10 [PI-Cer(d36:1)+H]+ [PS(36:0(OH))+H]+ 808.5678 808.569826 C42H83NO11P + -2.5057 16313 [PI-Cer(t36:0)+H-H2O] + [PG(36:2(OH))+NH4] + 808.5802 808.578068 C42H82NO13 [LacCer(d30:0)+H] 2.6367 1669 + 808.6631 808.663690 C46H91NO8Na [HexCer(d40:0)+Na] -0.7296 333 809.5584 not matched 27 [PG(O-40:6)+H]+ [PG(P-40:5)+H]+ + [PG(40:4)+H-H2O] 809.5709 809.569097 C46H82O9P [PG(O-40:5(OH))+H- 2.2271 12401 + H2O] [PG(P-40:4(OH))+H- + H2O] [PA(O-42:2)+K]+ 809.5834 809.582100 C H O PK 1.6058 559 45 87 7 [PA(P-42:1)+K]+ + 809.6475 809.650696 C45H91N2O6PNa [SM(d40:1)+Na] -3.9474 12 809.6662 not matched 47 + 810.3835 810.380049 C33H66NO17P2 [PIP(24:1(OH))+NH4] 4.2585 16 + 810.5741 810.575960 C42H84NO11S [SHexCer(d36:0)+H] -2.2947 3191

Table 6.1 (cont.) 207

Molecular Error Number Measured m/z Exact m/z Assignment Formula (ppm) of Cells [PC(36:1)+Na]+ + 810.5991 810.600732 C46H85NO8P [PC(O-36:2(OH))+Na] -2.0133 14060 [PC(P-36:1(OH))+Na]+ [HexCer(d42:2)+H- H O]+ 810.6788 810.681745 C H NO 2 -3.6327 369 48 92 8 [HexCer(t42:1)+H- + H2O] [PA(O-42:1)+K]+ 811.5992 811.597750 C H O PK 1.7866 4774 45 89 7 [PA(P-42:0)+K]+ 811.6832 not matched 59 + [PA(O-46:1)+H-H2O] 811.6929 811.694020 C49H96O6P + -1.3798 48 [PA(P-46:0)+H-H2O] + 811.7273 811.730253 C55H96O2Na [CE(28:2)+Na] -3.6379 13 812.5383 not matched 184 + 812.6024 812.601241 C44H87NO9K [HexCer(t38:0)+K] 1.4263 559 [PC(36:0)+Na]+ + 812.6133 812.613976 C44H88NO8PNa [PC(O-36:1(OH))+Na] -0.8319 268 [PC(P-36:0(OH))+Na]+ [SM(d42:2)+H]+ 813.6805 813.684402 C47H94N2O6P + -4.7955 1254 [SM(t42:1)+H-H2O] 814.4150 not matched 860 814.4884 not matched 4021 [PE(42:9)+H]+ [PE(42:8(OH))+H- 814.5384 814.538131 C47H77NO8P + 0.3302 145 H2O] + [PA(44:10)+NH4] + 814.6290 814.629626 C44H90NO8PNa [PC(O-36:0(OH))+Na] -0.7684 712 814.6853 not matched 562 815.4260 not matched 6900 815.4916 not matched 2216 PI(P-34:4)+H]+ LPI(34:5)+H]+ + [PI(34:3)+H-H2O] 815.5104 815.506891 C43H76O12P [PI(O-34:4(OH))+H- 4.3028 64 + H2O] [PI(P-34:3(OH))+H- + H2O] 815.6338 not matched 126 [SM(d42:1)+H]+ 815.6979 815.700052 C47H96N2O6P + -2.6382 205 [SM(t42:0)+H-H2O] 816.4292 not matched 1004

Table 6.1 (cont.) 208

Molecular Error Number Measured m/z Exact m/z Assignment Formula (ppm) of Cells [PC(38:1)+H]+ [PC(O-38:2(OH))+H]+ [PC(P-38:1(OH))+H]+ 816.6448 816.647682 C H NO P -3.5291 1724 46 91 8 [PC(38:0(OH))+H- + H2O] + [PS(O-40:0)+H-H2O] 816.7011 not matched 21 817.3574 not matched 170 817.6480 not matched 565 818.3699 not matched 218 819.3747 not matched 32 820.3873 not matched 704 + 820.5279 820.527567 C49H75NO7P [PE(44:11)+H-H2O] 0.4058 13 821.3889 not matched 240 + [CerP(t46:0)+NH4] 821.7092 821.710617 C46H98N2O7P + -1.7245 14 [PC(O-38:0)+NH4] + 822.6405 822.642954 C46H89NO9Na [HexCer(t40:1)+Na] -2.9831 681 + 823.4515 823.452208 C42H73O11PK [PG(36:5(OH))+K] -0.8598 247 + 823.5781 823.580725 C42H83N2O11P [PI-Cer(d36:2)+NH4] -3.1873 11 823.6438 not matched 194 + 824.5532 824.555224 C42H82NO12S [SHexCer(t36:1)+H] -2.4547 13 + 825.4665 825.467858 C42H75O11PK [PG(36:4(OH))+K] -1.6451 107 825.5566 not matched 50 826.4697 not matched 24 + 826.5691 826.570875 C42H84NO12S [SHexCer(t36:0)+H] -2.1474 332 827.5723 not matched 44 + 827.6586 827.661261 C45H93N2O7PNa [SM(t40:0)+Na] -3.2151 10 828.3937 not matched 12731 [PC(40:9)+H]+ 828.5549 828.553782 C48H79NO8P [PC(40:8(OH))+H- 1.3493 598 + H2O] 829.3969 not matched 3277 + [PI(O-36:3)+H-H2O] 829.5581 829.558926 C45H82O11P + -0.9957 82 [PI(P-36:2)+H-H2O] 830.4845 not matched 6450 + 830.5126 830.509663 C45H78NO8PK [PE(40:6)+K] 3.5364 10 [PC(O-38:6)+K]+ 830.5482 830.546049 C H NO PK 2.5899 3264 46 82 7 [PC(P-38:5)+K]+ [LacCer(d32:3)+H]+ 830.5632 830.562418 C44H80NO13 [LacCer(t32:2)+H- 0.9415 26 + H2O]

Table 6.1 (cont.)

209

Molecular Error Number Measured m/z Exact m/z Assignment Formula (ppm) of Cells 831.4877 not matched 2114 + 831.5533 831.553447 C48H80O9P [PG(42:7)+H-H2O] -0.1768 873 [PC(40:7)+H]+ + 832.5865 832.585082 C48H83NO8P [PS(O-42:6)+H-H2O] 1.7031 1402 + [PS(P-42:5)+H-H2O] + 832.6615 832.663690 C48H91NO8Na [HexCer(d42:2)+Na] -2.6301 1666 833.3535 not matched 167 + 833.3704 833.372029 C41H62O15K [DGDG(26:7)+K] -1.9547 57 [PG(O-40:5)+Na]+ 833.5635 833.566692 C H O PNa -3.8293 411 46 83 9 [PG(P-40:4)+Na]+ + [PI(O-36:1)+H-H2O] 833.5897 833.590227 C45H86O11P + -0.6322 446 [PI(P-36:0)+H-H2O] + 833.6647 833.662998 C47H94O9P [PA(44:0(OH))+H] 2.0416 1123 [SM(t42:0)+H]+ + 833.7079 833.710617 C47H98N2O7P [PE(O-42:1)+NH4] -3.2589 30 + [PE(P-42:0)+NH4] 834.3604 not matched 287 [PE(42:7(OH))+H]+ 834.5667 834.564346 C47H81NO9P + 2.8206 79 [PA(44:8(OH))+NH4] [PI-Cer(d38:2)+H]+ [PS(38:1(OH))+H]+ 834.5836 834.585476 C44H85NO11P + -2.2478 815 [PI-Cer(t38:1)+H-H2O] + [PG(38:3(OH))+NH4] [PC(40:6)+H]+ [PC(P-40:6(OH))+H]+ [PC(40:5(OH))+H- 834.6042 834.600732 C48H85NO8P + 4.1553 4787 H2O] + [PS(O-42:5)+H-H2O] + [PS(P-42:4)+H-H2O] + 834.6773 834.679340 C48H93NO8Na [HexCer(d42:1)+Na] -2.4440 480 834.7111 not matched 12 835.3637 not matched 34 [PG(O-40:4)+Na]+ 835.5793 835.582342 C H O PNa -3.6406 19 46 85 9 [PG(P-40:3)+Na]+ + 835.6074 835.605877 C45H88O11P [PI(O-36:0)+H-H2O] 1.8226 1358 + 835.6637 835.666346 C47H93N2O6PNa [SM(d42:2)+Na] -3.1663 623 835.6806 not matched 82 836.3819 not matched 166 836.3969 not matched 27

Table 6.1 (cont.)

210

Molecular Error Number Measured m/z Exact m/z Assignment Formula (ppm) of Cells [PI-Cer(d38:1)+H]+ [PS(38:0(OH))+H]+ 836.5994 836.601126 C44H87NO11P + -2.0631 9862 [PI-Cer(t38:0)+H-H2O] + [PG(38:2(OH)+NH4] [PC(38:2)+Na]+ + 836.6144 836.613976 C46H88NO8PNa [PC(O-38:3(OH))+Na] 0.5068 194 [PC(P-38:2(OH))+Na]+ 836.6669 not matched 260 837.3663 not matched 461 837.3851 not matched 22 + 837.5014 837.502474 C44H74N2O11P [PS(38:8(OH))+NH4] -1.2824 14 [PG(O-42:6)+H]+ [PG(P-42:5)+H]+ + [PG(42:4)+H-H2O] 837.6026 837.600397 C48H86O9P [PG(O-42:5(OH))+H- 2.6301 5418 + H2O] [PG(P-42:4(OH))+H- + H2O] + 837.6664 837.669146 C45H94N2O9P [PE(40:0(OH))+NH4] -3.2781 45 + 837.6795 837.681996 C47H95N2O6PNa [SM(d42:1)+Na] -2.9797 37 838.3696 not matched 137 + 838.5027 838.501746 C48H73NO9P [PS(42:10)+H-H2O] 1.1377 1050 + 838.6059 838.607260 C44H88NO11S [SHexCer(d38:0)+H] -1.6217 287 + 838.6265 838.629626 C46H90NO8PNa [PC(38:1)+Na] -3.7275 437 839.4459 not matched 20 + [PI(36:5)+H-H2O] 839.5078 839.506891 C45H76O12P [PI(P-36:5(OH))+H- 1.0828 44 + H2O] + 839.5697 839.566124 C42H83N2O12S [SHexCer(t36:2)+NH4] 4.2593 12 [PA(O-44:1)+K]+ 839.6297 839.629050 C H O PK 0.7742 57 47 93 7 [PA(P-44:0)+K]+ 841.4618 not matched 415 841.4805 not matched 31 + 841.5630 841.558926 C46H82O11P [PG(40:5(OH))+H] 4.8410 10 + 841.5762 841.580056 C43H86O13P [PI(O-34:0(OH))+H] -4.5818 32 842.4181 not matched 94 842.4650 not matched 10 + 842.5212 842.523627 C42H77NO14Na [LacCer(t30:2)+Na] -2.8806 2820 + 842.6600 842.660926 C46H94NO8PNa [PC(O-38:0(OH))+Na] -1.0989 279 + 843.3688 843.369150 C37H65O17P2 [PIP(28:4(OH))+H] -0.4150 62 + 843.5264 843.528295 C47H76N2O9P [PE(42:11(OH))+NH4] -2.2465 856

Table 6.1 (cont.)

211

Molecular Error Number Measured m/z Exact m/z Assignment Formula (ppm) of Cells 843.6651 not matched 31 + 844.5296 844.531034 C42H80NO12PNa [PI-[Cer(t36:2)+Na] -1.6980 11 [PC(40:1)+H]+ [PC(O-40:2(OH))+H]+ 844.6777 844.678982 C H NO P -1.5177 875 48 95 8 [PC(P-40:1(OH))+H]+ + [PS(O-42:0)+H-H2O] 845.6810 not matched 277 846.4817 not matched 89 [PI-Cer(d36:1)+K]+ 846.5248 846.525707 C H NO PK -1.0714 10 42 82 11 [PS(36:0(OH))+K]+ + 848.5613 848.562334 C42H84NO12PNa [PI-Cer(t36:0)+Na] -1.2185 13 [PE(O-42:4)+K]+ 848.5969 848.592999 C H NO PK 4.5970 95 47 88 7 [PE(P-42:3)+K]+ + 848.6569 848.658604 C48H91NO9Na [HexCer(t42:2)+Na] -2.0079 1264 [PG(42:7)+H]+ 849.5645 849.564012 C48H82O10P [PG(42:6(OH))+H- 0.5744 12 + H2O] + 849.6002 849.600397 C49H86O9P [PA(46:6(OH))+H] -0.2319 45 849.6602 not matched 878 850.3990 not matched 17 850.6578 not matched 18 + 850.6728 850.674254 C48H93NO9Na [HexCer(t42:1)+Na] -1.7092 1661 851.6760 not matched 707 + [PIP(28:0)+NH4] 852.4617 852.463385 C37H76NO16P2 [PIP(P- -1.9766 81 + 28:0(OH))+NH4] [PC(O-42:4)+H]+ [PC(P-42:3)+H]+ + [PC(42:2)+H-H2O] 852.6868 852.684067 C50H95NO7P [PC(O-42:3(OH))+H- 3.2052 27 + H2O] [PC(P-42:2(OH))+H- + H2O] 853.3619 not matched 225 [PG(40:2)+Na]+ + 853.5906 853.592906 C46H87O10PNa [PG(O-40:3(OH))+Na] -2.7015 29 [PG(P-40:2(OH))+Na]+ 854.3651 not matched 24 + [PI(34:0)+NH4] + 856.5891 856.590955 C43H87NO13P [PI(O-34:1(OH))+NH4] -2.1656 278 + [PI(P-34:0(OH))+NH4] + 856.6529 856.655599 C47H96NO7PK [PE(O-42:0)+K] -3.1506 42

Table 6.1 (cont.) 212

Molecular Error Number Measured m/z Exact m/z Assignment Formula (ppm) of Cells 857.3861 not matched 13 + [PI(O-38:3)+H-H2O] 857.5924 857.590227 C47H86O11P + 2.5338 29 [PI(P-38:2)+H-H2O] 857.6561 not matched 10 858.4156 not matched 17 858.4794 not matched 14 [PS(42:9)+H]+ [PS(42:8(OH))+H- 858.5319 858.527961 C48H77NO10P + 4.5881 236 H2O] + [PG(42:11)+NH4] [PI-Cer(d38:1)+Na]+ 858.5825 858.583070 C H NO PNa -0.6639 391 44 86 11 [PS(38:0(OH))+Na]+ [PE(44:1)+H]+ [PE(O-44:2(OH))+H]+ [PE(P-44:1(OH))+H]+ [PA(46:2)+NH ]+ 858.6931 858.694632 C H NO P 4 -1.7841 13 49 97 8 [PA(O- + 46:3(OH))+NH4] [PA(P- + 46:2(OH))+NH4] 859.4020 not matched 216 [PI(36:4)+H]+ [PI(O-36:5(OH))+H]+ 859.5351 859.533106 C H O P 2.3199 71 45 80 13 [PI(P-36:4(OH))+H]+ + [PI(36:3(OH))+H-H2O] + 859.5839 859.584747 C50H84O9P [PG(44:7)+H-H2O] -0.9854 62 860.4090 not matched 100 + 860.4990 860.495456 C42H79NO12SK [SHexCer(t36:2)+K] 4.1186 4185 + 861.5023 861.502474 C46H74N2O11P [PS(40:10(OH))+NH4] -0.2020 1746 + 861.6673 861.669146 C47H94N2O9P [PE(42:2(OH))+NH4] -2.1423 134 [PE(O-46:6)+H]+ [PE(P-46:5)+H]+ + [PE(46:4)+H-H2O] [PE(O-46:5(OH))+H- 862.6705 862.668417 C51H93NO7P + 2.4146 38 H2O] [PE(P-46:4(OH))+H- + H2O] + [PA(P-48:6)+NH4] [PI-Cer(d40:1)+H]+ + 864.6321 864.632426 C46H91NO11P [PS(40:0(OH))+H] -0.3770 85 + [PI-Cer(t40:0)+H-H2O]

Table 6.1 (cont.)

213

Molecular Error Number Measured m/z Exact m/z Assignment Formula (ppm) of Cells [PC(40:2)+Na]+ + 864.6452 864.645276 C48H92NO8PNa [PC(O-40:3(OH))+Na] -0.0879 9 [PC(P-40:2(OH))+Na]+ + 866.5355 866.533046 C50H77NO9P [PS(44:10)+H-H2O] 2.8320 893 [PC(40:1)+Na]+ + 866.6611 866.660926 C48H94NO8PNa [PC(O-40:2(OH))+Na] 0.2008 182 [PC(P-40:1(OH))+Na]+ 866.6742 not matched 25 [PI(P-38:6)+H]+ [PI(38:5+H-H O]+ 867.5387 867.538191 C H O P 2 0.5867 105 47 80 12 [PI(P-38:5(OH))+H- + H2O] + 867.6306 867.626334 C55H88O5K DG(52:10)+K] 4.9169 5047 [PA(O-46:1)+K]+ 867.6645 867.660350 C H O PK 4.7830 15 49 97 7 [PA(P-46:0)+K]+ 868.5776 not matched 67 868.6339 not matched 2300 + 869.5884 869.587821 C46H87O11PNa [PG(40:2(OH))+Na] 0.6658 39 + 869.6109 869.611356 C45H90O13P [PI(O-36:0(OH))+H] -0.5244 31 870.3816 not matched 36 870.4435 not matched 62 + 870.5016 870.504578 C47H78NO9PK [PE(42:8(OH))+K] -3.4210 45 870.5935 not matched 12428 [PC(42:2)+H]+ [PC(O-42:3(OH))+H]+ [PC(P-42:2(OH))+H]+ 870.6948 870.694632 C50H97NO8P [PC(42:1(OH))+H- 0.1929 598 + H2O] + [PS(O-44:1)+H-H2O] + [PS(P-44:0)+H-H2O] + 871.4524 871.452208 C46H73O11PK [PG(40:9(OH))+K] 0.2203 19 + 871.5968 871.595981 C50H84N2O8P [PC(42:10)+NH4] 0.9396 11114 871.6981 not matched 203 872.3975 not matched 47 872.4125 not matched 183 872.6001 not matched 5936 [PE(44:5)+Na]+ + 872.6132 872.613976 C49H88NO8PNa [PE(O-44:6(OH))+Na] -0.8893 51 [PE(P-44:5(OH))+Na]+

Table 6.1 (cont.)

214

Molecular Error Number Measured m/z Exact m/z Assignment Formula (ppm) of Cells [PC(42:1)+H]+ [PC(O-42:2(OH))+H]+ 872.7107 872.710282 C H NO P 0.4790 990 50 99 8 [PC(P-42:1(OH))+H]+ + [PS(O-44:0)+H-H2O] 873.3727 not matched 212 + 873.4196 873.416100 C39H71O17P2 [PIP(30:3(OH))+H] 4.0072 273 + 873.6033 873.600397 C51H86O9P [PA(48:8(OH))+H] 3.3230 21 873.7140 not matched 473 [PA(48:0)+H]+ + 873.7271 873.730683 C51H102O8P [PA(O-48:1(OH))+H] -4.1008 52 [PA(P-48:0(OH))+H]+ 874.3816 not matched 154 874.4172 not matched 1083 874.4360 not matched 111 + 874.5166 874.520470 C46H78NO11PNa [PS(40:6(OH))+Na] -4.4253 15 874.7173 not matched 12 875.3861 not matched 25 875.4218 not matched 776 876.4288 not matched 409 876.4475 not matched 79 + 876.4963 876.494013 C49H76NO8PK [PE(44:11)+K] 2.6093 74 + 876.5957 876.593635 C44H88NO12PNa [PI-Cer(t38:0)+Na] 2.3557 1 [PE(O-44:4)+K]+ 876.6276 876.624299 C H NO PK 3.7656 10 49 92 7 [PE(P-44:3)+K]+ + 877.4433 877.447401 C39H75O17P2 [PIP(30:1(OH))+H] -4.6738 200 [PG(44:7)+H]+ 877.5952 877.595312 C50H86O10P [PG(44:6(OH))+H- -0.1276 87 + H2O] + 877.725598 C50H102O9P [PG(O-44:0)+H] 1.0276 877.7265 + 14 877.725562 C55H98O6Na [TG(52:4)+Na] 1.0687 878.4335 not matched 22 879.4874 not matched 17 + 879.741212 C55H100O6Na [TG(52:3)+Na] 1.3504 879.7424 + 14 879.743617 C57H99O6 [TG(54:6)+H] -1.3833 880.3913 not matched 16 880.6482 not matched 41 880.7457 not matched 11 + 881.7584 881.759267 C57H101O6 [TG(54:5)+H] -0.9833 15 882.4841 not matched 59 + 882.5723 882.573554 C46H85NO11SNa [SHexCer(d40:3)+Na] -1.4208 194

Table 6.1 (cont.)

215

Molecular Error Number Measured m/z Exact m/z Assignment Formula (ppm) of Cells + 882.6323 882.633475 C46H92NO12S [SHexCer(t40:0)+H] -1.3312 52 [PG(42:1)+Na]+ + 883.6356 883.639857 C48H93O10PNa [PG(O-42:2(OH))+Na] -4.8176 18 [PG(P-42:1(OH))+Na]+ 886.4410 not matched 26 887.4012 not matched 115 888.4083 not matched 94 + 888.5339 888.536272 C44H84NO12PK [PI-Cer(t38:2)+K] -2.6696 3126 889.4172 not matched 168 889.4528 not matched 20 [PG(42:6)+K]+ 889.5372 889.535544 C H O PK 1.8616 804 48 83 10 [PG(P-42:6(OH))+K]+ + 889.7116 889.713296 C51H99N2O6PNa [SM(d46:3)+Na] -1.9062 10 890.4261 not matched 163 891.4331 not matched 124 + 892.5790 892.582676 C51H84NO8PNa [PE(46:9)+Na] -4.1184 47 [PC(44:5)+H]+ [PC(O-44:6(OH))+H]+ + 892.6784 892.678982 C52H95NO8P [PC(P-44:5(OH))+H] -0.6520 48 + [PS(O-46:4)+H-H2O] + [PS(P-46:3)+H-H2O] + 894.5969 894.598326 C51H86NO8PNa [PE(46:8)+Na] -1.5940 138 [PC(44:4)+H]+ [PC(O-44:5(OH))+H]+ [PC(P-44:4(OH))+H]+ 894.6944 894.694632 C52H97NO8P [PC(44:3(OH))+H- -0.2593 186 + H2O] + [PS(O-46:3)+H-H2O] + [PS(P-46:2)+H-H2O] + 895.6658 895.666498 C49H97N2O7PK [SM(t44:2)+K] -0.7793 34 895.6977 not matched 19 + 896.6147 896.613976 C51H88NO8PNa [PE(46:7)+Na] 0.8075 10641 + 896.6466 896.649125 C47H94NO12S [SHexCer(t41:0)+H] -2.8160 29 + 897.6180 897.619121 C48H91O11PNa [PG(42:2(OH))+Na] -1.2489 8816 [PE(46:6)+Na]+ 898.6288 898.629626 C H NO PNa -0.9192 8815 51 90 8 [PE(P-46:6(OH))+Na]+ + 898.7282 898.728302 C59H96NO5 [DG(56:12)+NH4] -0.1135 16 + 899.6322 899.634771 C48H93O11PNa [PG(42:1(OH))+Na] -2.8578 6380

Table 6.1 (cont.)

216

Table 6.1 (cont.) 858 lipid features detected by high throughput single cell FT-ICR MALDI MS in 10 or more cells, their mass matched putative assignments, and number of cells where they are observed. Only lipids known to be present in mammalian tissues are used for lipid assignment.

217

6.7 References

(1) S. Herculano-Houzel, Frontiers in Human Neuroscience 2009, 3, 31.

(2) V. Matyash, H. Kettenmann, Brain Research Reviews 2010, 63, 2-10.

(3) L. J. Lawson, V. H. Perry, P. Dri, S. Gordon, Neuroscience 1990, 39, 151-170.

(4) C. Schitine, L. Nogaroli, M. R. Costa, C. Hedin-Pereira, Frontiers in Cellular

Neuroscience 2015, 9.

(5) D. J. Field, R. A. Collins, J. C. Lee, Proceedings of the National Academy of Sciences

1984, 81, 4041-4045.

(6) W. Eric Thomas, Brain Research Reviews 1992, 17, 61-74.

(7) C. A. Colton, Journal of Neuroimmune Pharmacology 2013, 8, 145-162.

(8) E. Bullmore, O. Sporns, Nature Reviews Neuroscience 2009, 10, 186.

(9) M. Schachner, in Neuroimmunology (Ed.: J. Brockes), Springer US, Boston, MA, 1982,

pp. 215-250.

(10) O. Stegle, S. A. Teichmann, J. C. Marioni, Nature Reviews Genetics 2015, 16, 133.

(11) B. M. Evers, C. Rodriguez-Navas, R. J. Tesla, J. Prange-Kiel, C. R. Wasser, K. S. Yoo, J.

McDonald, B. Cenik, T. A. Ravenscroft, F. Plattner, R. Rademakers, G. Yu, C. L. White,

J. Herz, Cell Reports 2017, 20, 2565-2574.

(12) A. A. Spector, M. A. Yorek, Journal of Lipid Research 1985, 26, 1015-1035.

(13) J. Jumpsen, M. T. Clandinin, Brain development: relationship to dietary lipid and lipid

metabolism, AOCS Publishing, 1995.

(14) R. M. Adibhatla, J. F. Hatcher, in Lipids in Health and Disease (Eds.: P. J. Quinn, X.

Wang), Springer Netherlands, Dordrecht, 2008, pp. 241-268.

(15) M. R. Wenk, Cell 2010, 143, 888-895.

218

(16) L. Li, R. W. Garden, J. V. Sweedler, Trends in Biotechnology 2000, 18, 151-160.

(17) T.-H. Ong, D. J. Kissick, E. T. Jansson, T. J. Comi, E. V. Romanova, S. S. Rubakhin, J.

V. Sweedler, Analytical Chemistry 2015, 87, 7036-7042.

(18) E. T. Jansson, T. J. Comi, S. S. Rubakhin, J. V. Sweedler, ACS Chemical Biology 2016,

11, 2588-2595.

(19) T. D. Do, J. F. Ellis, E. K. Neumann, T. J. Comi, E. G. Tillmaand, A. E. Lenhart, S. S.

Rubakhin, J. V. Sweedler, ChemPhysChem 2018, 19, 1180-1191.

(20) T. J. Comi, E. K. Neumann, T. D. Do, J. V. Sweedler, Journal of the American Society

for Mass Spectrometry 2017, 28, 1919-1928.

(21) H. Xianlin, G. R. W., Mass Spectrometry Reviews 2005, 24, 367-412.

(22) K. Sandra, A. d. S. Pereira, G. Vanhoenacker, F. David, P. Sandra, Journal of

Chromatography A 2010, 1217, 4087-4099.

(23) N. Goto-Inoue, T. Hayasaka, N. Zaima, M. Setou, Biochimica et Biophysica Acta (BBA) -

Molecular and Cell Biology of Lipids 2011, 1811, 961-969.

(24) M. Kompauer, S. Heiles, B. Spengler, Nature Methods 2016, 14, 90.

(25) M. Sud, E. Fahy, D. Cotter, E. A. Dennis, S. Subramaniam, Journal of Chemical

Education 2012, 89, 291-292.

(26) E. Fahy, M. Sud, D. Cotter, S. Subramaniam, Nucleic Acids Research 2007, 35, W606-

W612.

(27) T. Harayama, H. Riezman, Nature Reviews Molecular Cell Biology 2018, 19, 281.

(28) A. Shevchenko, K. Simons, Nature Reviews Molecular Cell Biology 2010, 11, 593.

(29) D. Piomelli, G. Astarita, R. Rapaka, Nature Reviews Neuroscience 2007, 8, 743.

(30) R. H. Masland, Current Biology 2004, 14, R497-R500.

219

(31) G. M. Shepherd, The synaptic organization of the brain, Vol. 198, oxford university press

New York, 1998.

(32) S. Miyawaki, H. Imai, T. Hayasaka, N. Masaki, H. Ono, T. Ochi, A. Ito, H. Nakatomi, M.

Setou, N. Saito, Neuroscience 2016, 322, 66-77.

(33) S. N. Jackson, H.-Y. J. Wang, A. S. Woods, Journal of the American Society for Mass

Spectrometry 2005, 16, 2052-2056.

(34) A. S. Mohammadi, N. T. N. Phan, J. S. Fletcher, A. G. Ewing, Analytical and

Bioanalytical Chemistry 2016, 408, 6857-6868.

(35) K. Sandhoff, Biochimie 2016, 130, 146-151.

(36) S. Grassi, E. Chiricozzi, L. Mauri, S. Sonnino, A. Prinetti, Journal of Neurochemistry, 0.

(37) K. Bienias, A. Fiedorowicz, A. Sadowska, S. Prokopiuk, H. Car, Pharmacological

Reports 2016, 68, 570-581.

(38) T. Sural-Fehr, E. R. Bongarzone, Journal of Neuroscience Research 2016, 94, 1042-

1048.

(39) A. S. Woods, S. N. Jackson, The AAPS Journal 2006, 8, E391-E395.

(40) A. Veloso, R. Fernández, E. Astigarraga, G. Barreda-Gómez, I. Manuel, M. T. Giralt, I.

Ferrer, B. Ochoa, R. Rodríguez-Puertas, J. A. Fernández, Analytical and Bioanalytical

Chemistry 2011, 401, 89-101.

(41) L. S. Eberlin, C. R. Ferreira, A. L. Dill, D. R. Ifa, R. G. Cooks, Biochimica et Biophysica

Acta (BBA) - Molecular and Cell Biology of Lipids 2011, 1811, 946-960.

(42) T. J. Comi, M. A. Makurath, M. C. Philip, S. S. Rubakhin, J. V. Sweedler, Analytical

Chemistry 2017, 89, 7765-7772.

220

(43) E. K. Neumann, T. J. Comi, N. Spegazzini, J. W. Mitchell, S. S. Rubakhin, M. U.

Gillette, R. Bhargava, J. V. Sweedler, Analytical Chemistry 2018, 90, 11572-11580.

(44) S. R. Fagerer, T. Schmid, A. J. Ibáñez, M. Pabst, R. Steinhoff, K. Jefimovs, P. L. Urban,

R. Zenobi, Analyst 2013, 138, 6732-6736.

(45) B. J. Andreone, B. W. Chow, A. Tata, B. Lacoste, A. Ben-Zvi, K. Bullock, A. A. Deik,

D. D. Ginty, C. B. Clish, C. Gu, Neuron 2017, 94, 581-594.e585.

(46) R. P. Bazinet, S. Layé, Nature Reviews Neuroscience 2014, 15, 771.

(47) C. Betsholtz, Nature 2014, 509, 432.

(48) L. v. d. Maaten, G. Hinton, Journal of machine learning research 2008, 9, 2579-2605.

(49) K. Shekhar, S. W. Lapan, I. E. Whitney, N. M. Tran, E. Z. Macosko, M. Kowalczyk, X.

Adiconis, J. Z. Levin, J. Nemesh, M. Goldman, S. A. McCarroll, C. L. Cepko, A. Regev,

J. R. Sanes, Cell 2016, 166, 1308-1323.e1330.

(50) V. D. Blondel, J.-L. Guillaume, R. Lambiotte, E. Lefebvre, Journal of statistical

mechanics: theory and experiment 2008, 2008, P10008.

221

CHAPTER 7

MULTIMODAL CHEMICAL ANALYSIS OF THE BRAIN BY HIGH RESOLUTION MASS

SPECTROMETRY AND INFRARED SPECTROSCOPIC IMAGING

Notes and Acknowledgements

This chapter is adapted from a completed manuscript coauthored by Troy J. Comi,

Nicolas Spegazzini, Jennifer W. Mitchell, Stanislav S. Rubakhin, Martha U. Gillette, Rohit

Bhargava, and Jonathan V. Sweedler, submitted for publication on June 27, 2018 and accepted

September 6, 2018, in Analytical Chemistry, DOI: 10.1021/acs.analchem.8b02913. T.J. Comi performed most data analysis and write all scripts used in the study. N. Spegazzini performed the

QCL-IR analysis. S.S. Rubakhin, M.U. Gillette, R. Bhargava, and J.V. Sweedler assisted with concept development and writing, as well as funded the work. We acknowledge Dr. Jennifer

Mitchell and Joseph F. Ellis for useful discussions. The authors gratefully acknowledge support from the National Institutes of Health, Award Number P30 DA018310 from the National

Institute on Drug Abuse, and from the National Institute of Mental Health, Award Number 1U01

MH109062. E.K. Neumann and T.J. Comi acknowledge support from the National Science

Foundation Graduate Research Fellowship Program and the Springborn Fellowship. T.J. Comi received additional support through the Training Program at Chemistry-Interface with Biology

(T32 GM070421). This work is also partially supported by the Agilent Thought Leader Award to

R. Bhargava.

7.1 Introduction

The brain is a complex organ that is incompletely understood, partly due to its vast chemical and cellular heterogeneity. Communication between different cell types via the exchange of chemical signals contributes to emergent functions, including episodic and spatial

222 memory, in which the hippocampus plays a critical role.[1,2] While many morphological regions of the hippocampus are involved in the formation and maintenance of different memory types, the dentate gyrus (DG) is implicated in the formation of spatial and long-term memory.[3,4] A wide range of distinct cell types are present in the DG, including granule cells, astrocytes, ependyma, radial glia, neuroblasts, and inhibitory neurons.[5] Several analytical approaches have effectively been used to advance our understanding of brain function, e.g., electrophysiology,[6] immunohistochemistry,[7] behavioral analysis,[8] magnetic resonance imaging,[9] among others.[10-

13] Despite this progress, there remains a need for techniques capable of untargeted, multiplex chemical analysis at spatial resolutions relevant to cellular length scales.

Matrix-assisted laser desorption/ionization (MALDI) mass spectrometry imaging (MSI) is capable of multiplex detection and structural characterization of hundreds to thousands of analytes within a sample at femtomole detection limits.[14] However, MALDI MSI typically obtains spatial resolutions greater than 20 µm,[15] precluding subcellular imaging of tissue sections using most commercial instruments. Several labs have built prototype MALDI mass spectrometers capable of subcellular resolution,[16,17] or used oversampling techniques on commercial instruments[18] to achieve smaller pixel sizes. However, these approaches can reduce chemical coverage due to the decreased absolute number of analytes present in a single pixel of the chemical image.

Although instrument modifications are one approach to achieving higher spatial resolutions, an alternative is to enhance the resolution of mass spectrometry (MS) ion images through sharpening using another imaging modality.[19-21] When an ion distribution image is correlated to an underlying, high spatial resolution image, the results of sharpening are encouraging. However, generally only a single, high spatial resolution image is available for

223 sharpening. Pan sharpening ion images with contrasting high spatial resolution images can induce artifacts or degrade the quality of the original MS image. Therefore, a singular, high spatial resolution image limits the number and diversity of ion images suitable for sharpening, possibly excluding important analytes. Here, we overcome this limitation by sharpening MS images with those obtained using another chemical imaging approach.

Infrared (IR) spectroscopic imaging is a non-destructive, optical imaging method that provides rich spectral information at micron spatial resolution.[22,23] Numerical algorithms can relate the data to histologic identity or transformation of constituent cells,[24-28] and provide images that mimic those obtained from conventional staining protocols.[29,30] The collected data can be used to identify classes of chemicals in tissue[31] but not most individual metabolites in complex biological samples. Thus, coupling IR and MSI presents a unique opportunity to integrate complementary spectral data and spatial resolutions. We applied this combined approach to measure chemical distributions present within the hippocampus, including the DG.

Because of its promise, the integration of MSI and vibrational spectroscopy has been a long-term goal.[32-34] Practitioners have recorded IR imaging data prior to MSI by combining synchrotron Fourier transform (FT)-IR microspectroscopy with time-of-flight (TOF)-secondary ion mass spectrometry to study histopathological changes in hepatic steatosis.[35,36] Recently, hyperspectral imaging was performed with FT-IR spectroscopy, confocal Raman spectroscopy, and MALDI MSI on hamster brains.[37] In that work, the FT-IR and Raman spectroscopic analyses were conducted on the same tissue section, and MALDI MSI was used to study an adjacent section.

Our goal was to demonstrate enhanced registration and image correlations using both modalities to interrogate the same sample. The choice of recording IR data first seems intuitive

224 and straightforward. However, as IR spatial resolution increases, the image acquisition time also rises to at least several hours, over which analytes in the fresh tissue (as required for MSI) may degrade. While MALDI MS is generally considered destructive, prior reports have shown that only a fraction of the sample is consumed during the measurement process.[38] After single cell

MALDI MS, sufficient material remains for follow-up analyses.[39,40] In imaging applications, tissue sections are frequently stained following MSI.[41-45] The MALDI matrix 2,5- dihydroxybenzoic acid (DHB) is utilized industrially as an antioxidant[46,47] and the vacuum created inside the ion source reduces oxidation during analysis.[48] This may be one of the first explorations of using these two imaging modalities, at their native optimal configurations, in which MS is performed first. Given the persistence of diagnostic information in IR imaging, despite the influence of various preparation conditions,[25,49-51] it is likely that IR chemical signatures arise from robust structural features, and different regions of the brain are likely conserved post-MALDI analysis as well.

We developed a combined chemical imaging approach that allows broadened spatio- chemical characterization of a variety of biological samples via sequential analysis of specimens using MSI followed by IR imaging. While counter-intuitive in terms of analysis order, the methodology preserves the information acquired by both analytical modalities. The resulting multimodal imaging dataset can be easily spatially registered through affine transformations and is suitable for data fusion, including pan sharpening. Compared to prior approaches for sharpening MS ion images utilizing optical monochrome morphological information, we used

“chemical pan sharpening” as a means to leverage both the higher spatial resolution IR image and spectral content to optimally fit the ion distribution images. Finally, we determined that data fusion of the image sets enables discrimination between anatomically relevant brain regions.

225

7.2 Experimental

Chemicals.

All chemicals were purchased from Millipore Sigma (St. Louis, MO) and used without further purification unless otherwise specified.

Animals.

Ten- to twelve-week old LE/BluGill rats (University of Illinois at Urbana-Champaign) were used for all studies. Use of this inbred strain greatly reduces inter-experimental variation common to outbred animals and allows achievement of high statistical significance with small sample sizes. A dense genome scan was performed at a 10 cM interval between markers on

LE/BluGill progenitors. The results of this scan, performed by the Medical College of Wisconsin

Human and Molecular Genetics Center as part of the National Heart, Lung, and Blood Institute

(NHLBI) Programs for Genomic Applications (PGA) U01 HL66579, demonstrated that the colony is inbred, yielding one allele at each locus tested (http://pga.mcw.edu/pga- bin/strain_desc.cgi).

All animal-related experimental procedures were conducted at the University of Illinois at Urbana-Champaign under protocols approved by the Institutional Animal Care and Use

Committee under Animal Welfare Assurance number A3118-01. All animal care and experiments were conducted in full compliance with the principles and procedures outlined in the National Institutes of Health Guide for the Care and Use of Laboratory Animals.

Tissue Sectioning.

Coronal brain sections containing the hippocampus were prepared from adult LE/BluGill rats. Rats were sacrificed by decapitation 2 h after lights were turned on in the donor colony.

Brains were quickly removed and placed on dry ice crushed to a fine powder. Coronal brains

226 were sectioned (20 µm) by cryostat at -17°C (Microm HM550, Thermo Fisher Scientific,

Waltham, MA). Sections containing a similar hippocampal rostral-caudal plane from three separate animals were placed on the same low emission (low-E) glass slide (Kevley

Technologies, Chesterland, OH) and stored with Drierite desiccant at –80°C until further analysis. Placing the three biological replicates on a single slide reduces batch variations for IR analysis and correlated MS analysis.

Sample Preparation and MS Analysis.

Tissue sections for MS analysis were warmed to room temperature in a dry nitrogen box for an hour before MALDI matrix application. DHB was sublimed onto tissue sections, forming a fine crystal layer with a thickness of ~0.3 mg/cm2 for lipid analysis using a custom sublimation chamber described previously.[52] Following sublimation, slides were returned to 22°C within a dry nitrogen box for at least 15 min. Prior to quantum cascade laser (QCL)-IR analysis (either before, after, or without MALDI MS analysis), sections were placed in a vacuum desiccator for 2 h.

Coated with DHB, hippocampal tissue sections were imaged at 1200 dpi with a flatbed scanner (Canon U.S.A. Inc., Melville, NY) to guide MSI. Optical images were loaded into flexImaging ver. 4.1 (Bruker Corp., Billerica, MA) and registered with a solariX XR 7T FT-ion cyclotron resonance (ICR) mass spectrometer (Bruker Corp.). Figure 7.1 displays the optimized protocol for correlated MALDI MSI and QCL-IR spectroscopy. Spectra were acquired in positive mode covering a mass range of m/z 150 to 3000, requiring a 0.7340 s transient. All MS data were acquired using the “minimum” laser probe setting (~25 µm diameter). The tissue was imaged with a 25 µm pixel size and two additional biological replicates were collected with 50

µm pixel sizes, with corresponding data presented in the supporting information. Data at each

227 pixel was generated by accumulating ions produced during ten laser shots at 25% power and

1000 Hz, unless otherwise specified. Reduced profile spectra were saved via FTMS control and a reduced file was generated within flexImaging for later import into SCiLS 2016b ver. 4.01.8720

(Bruker Corp). Within SCiLS, spectra were root mean square (RMS) normalized and exported in imzml format for further analysis.

Matrix Removal and Sample Fixation.

After MSI analysis, hippocampal tissue sections were placed within a solution of 4% paraformaldehyde (PFA) in 1× phosphate buffered saline (PBS; ThermoFisher Scientific) for 15 min to remove the MALDI matrix and concurrently fix the tissue sections. Tissues were washed in PBS for 3 min following fixation and briefly rinsed with Milli-Q water (Millipore, Billerica,

MA) to remove any remaining salts. Tissues were dried under a stream of nitrogen gas and placed within a desiccator for 2 h before QCL-IR imaging.

Infrared Analysis.

IR spectroscopic images were acquired using a prototype QCL-IR system (Agilent

Technologies, Santa Clara, CA). The discrete frequency QCL-IR system uses a room temperature single-point bolometer detector, with a 0.72 NA objective lens. Spectra were acquired in the mid-IR range, 1900–800 cm-1, at 8 cm-1 spectral resolution and 5 μm image pixel size in reflection mode. Spectral processing steps were implemented in MATLAB 2015b

(MathWorks, Inc., Natick, MA) and the ENVI-IDL 4.8 environment, and saved as .mat files for registration with MALDI images.

Data Registration and Analysis.

MSI datasets were imported into MATLAB as imzml files using a modified version of

MSIreader.[53] The series of mass spectra corresponding to each pixel with the MS image

228 consisted of m/z intensity pairs of centroid data. Mass spectra were aligned with non-uniform bin widths because mass resolution changes as a function of m/z value for FT-ICR MS. For example, the bin width at m/z 150 was 0.0004 Da and 0.05 Da at m/z 2500, with 10 additional divisions in between. Bin widths were estimated as the average peak width at m/z values over the range of the mass spectra. Bins were constructed as piecewise linear divisions over the entire spectral range.

Next, the centroided m/z values of each peak in the spectrum were counted and placed into finer bin divisions (one-eighth the values provided above) for outlier noise removal. The distribution of peak frequency was peak-picked with a minimum peak distance of the original bin width and a height threshold of 0.1% of the number of pixels. These operations removed peaks found in a small subset of pixels, producing a tentative m/z list for the aligned data set.

Next, the rough m/z list was refined by determining the center of mass for all peaks falling within the bin range. Thus, while binning was performed for alignment, the recorded m/z value retained high mass accuracy from the FT-ICR MS data. Finally, the signal intensity matrix was populated from the exact m/z list by summing the intensities within the recorded m/z value ± the bin width. The resulting data matrix was suitable for performing multivariate data analysis.

Image registration between MS and IR images was performed on the score image of principal component 1 (PC1) determined during principal component analysis (PCA) of the data matrix, which captures the majority of variance in each imaging mode. To speed computation, a subset of 1000 pixels was randomly chosen from each set of images to estimate the coefficient matrix by PCA. The entire score image was estimated by multiplying the signal intensity matrix by the coefficients determined with a subset of pixels. The score image resulting from PCA on a subset of pixels was qualitatively identical to the entire dataset and produced adequate alignment through image registration.

229

Next, the score images were roughly overlain by manually selecting control points from anatomical features present in each image. The initial, affine transformation was utilized to seed intensity-based registration, which produces better overlap. Manually determining the initial transformation was necessary to ensure consistent convergence of the intensity-based registration. The resulting affine transformation was then utilized to map the MS image onto the

IR image space.

Pan Sharpening.

Pan sharpening was performed with a Laplacian pyramid method, which utilizes high spatial frequency components from the higher spatial resolution IR image to sharpen chemical images obtained by MSI.[21] Briefly, this implementation incorporates the affine transformation from registration to estimate a scaling factor, and can match arbitrary differences in scale. First, the scale of the transformation is utilized to determine the number of iterations to down-sample the IR image. Each down-sample increases the pixel size by a factor of two. To match the image modalities for fusion, the MS image is up-sampled slightly to the next matching size, e.g., fusing the 5 µm IR pixels with the 25 µm MSI pixels required interpolating the MS image to 20 µm for two iterations of down-sampling.

At each iteration, the IR image is reduced with the kernel specified by Burt and Adelson, as implemented in the MATLAB function impyramid.[54] The difference between the effectively blurred reconstruction and the current image retains the high frequency features of the image, which is stored in the Laplacian pyramid. Once the IR image is resized to match the MS image, the process is repeated in reverse. At each iteration, the MS image is expanded by a factor of two and the Laplacian pyramid imparts the high frequency information from the IR image.

230

Midlevel Data Fusion.

Data fusion of each image set was performed by up-sampling the MS images to match IR image spatial resolution. Briefly, both MS and IR images were subjected to PCA for dimensionality reduction while retaining 99% of variance for the IR data and PC1–PC10 for the

MSI (~64–81% of variance). The choice of principal components attempted to balance the contributions of each modality while rejecting any unnecessary noise. To up-sample the MSI data, each score image was spatially transformed to the IR image with bilinear interpolation.

Data fusion was conducted pixel by pixel between the IR and MS images by concatenating the

PC scores and standardizing each principal component to be mean centered with unit variance.

The resulting, fused image was again subjected to PCA to identify the highest-variance contributions.

For unsupervised, multivariate classification, similar pixels were grouped together by k- means clustering for each IR spectral feature image, interpolated MS image, and fused score image. Within the hippocampus, clustering highlights anatomical regions that are chemically similar to each other. The optimum number of clusters was determined by Davies-Bouldin cluster evaluation. Statistically significant differences in ion intensity between anatomical regions were tested by one-way ANOVA with multiple comparison false discovery correction. A post-hoc Tukey-Karmer method was used for multiple comparison tests between significantly different ion intensities.

7.3 Results and Discussion

Optimized Acquisition of Multimodal Imaging Datasets.

Combining disparate analytical techniques frequently requires compromising established protocols to prepare a sample suitable for analysis by each method. The suboptimal performance

231 of a given method is an acceptable compromise if the orthogonal information acquired from the multimodal approach offsets the degraded data quality and/or analyte coverage compared to using a single method. To optimize multimodal data acquisition, a systematic assessment of

QCL-IR and MALDI MS analyses and their interactions was performed.

To analyze the same tissue section, some compromises between MALDI MS and IR imaging were required. IR imaging of tissue sections frequently employs salt plates with low IR absorbance, or low-E glass slides that reflect IR light effectively. Due to their lower cost, we utilized low-E glass slides for the IR acquisition. Initial attempts to increase low-E slide conductivity for MALDI-TOF analysis included gold sputtering following MALDI matrix application. This treatment introduced additional variance during sample preparation and left an insoluble residue after MALDI matrix removal, complicating subsequent IR analysis. Thus, we used a MALDI FT-ICR mass spectrometer, which functions appropriately with non-conductive substrates, likely because of the decoupling of analyte desorption/ionization and analysis, as well as the medium level of vacuum in the ion source.[55] We also found that spray-based MALDI matrix application led to unacceptable levels of analyte delocalization, as was evident in the resulting IR images. Thus, we applied MALDI matrix by sublimation, thereby improving IR imaging.

Since IR microspectroscopy is generally assumed to be non-destructive, in our initial experiments, IR imaging of tissue sections was conducted first. The post-IR imaging tissues were degraded in quality for further MSI measurements; many chemical compounds became undetectable and the overall signal intensity was lower compared to control tissue analyzed directly with MALDI MSI. To validate this finding, adjacent tissue sections were imaged by

MALDI MS with and without IR analysis (Figure 7.2). Optimum signal intensity was obtained

232 using MALDI MS with 25% laser power and 10 laser shots for the control sample (Figure

7.2A(i)). Using the same settings we were unable to acquire high quality mass spectra from tissue after IR imaging (Figure 7.2A(ii)). For example, the RMS-normalized intensity of the base peak in the average mass spectrum decreased from 1.6×107 to 4×106 counts following IR imaging (Figure 7.2B(i) and (ii)). As shown in Figure 7.2A(iii), doubling the laser power and number of laser shots were necessary to regain similar signal intensity (9×107 counts) but not the morphological detail of the ion images. While doubling the laser power increased the raw intensity above values obtained in the control measurement, the observed noise level also became disproportionally higher (~100 fold), demonstrating that the signal-to-noise ratio was still worse overall. Moreover, tissue exposed to higher MALDI MS laser power was physically damaged, preventing follow-up staining with hematoxylin and eosin, or immunohistochemistry.

Our finding that the tissue has degraded MS signal after IR imaging is counter intuitive as it disagrees with nano-IR studies which have shown minute temperature changes with QCL illumination. However, lipid degradation in unfixed tissue at ambient conditions has been shown to reduce MSI signal quality. Exploration of the mechanism and magnitude of tissue degradation is beyond the scope of this manuscript but is an area warranting further investigations. Thus, the approach of acquiring IR data first likely leads not only to poorer mass spectral data but may also provide additional variability in the IR measurement itself as the sample changes. One option to mitigate damage could be to use a tightly controlled protocol that may also involve nitrogen purging. But this was not explored here.

Next, we explored reversing the order of the multimodal measurements by first examining how MALDI MS alters the results of IR measurements. Adjacent hippocampus tissue sections were analyzed with IR imaging with various perturbations from the MALDI MSI

233 process. Tissues were either 1) directly fixed using PFA, 2) coated with DHB and then fixed, or

3) coated with DHB, analyzed by MALDI MS, and fixed. Figure 3A displays absorbance images

- -1 acquired at wavenumbers for the asymmetric PO2 stretch (1248 cm ) and amide II stretch (1656 cm-1). Qualitatively, the images indicate no damage from MALDI desorption and no apparent distortion or blurring of expected morphological image features. The average absorption spectrum was calculated from the hippocampal region, shown as shaded regions of interest, and plotted for direct comparison (Figure 7.3B). The spectra indicate a slight red shift in IR spectra collected from tissues exposed to DHB, with no change upon laser ablation. Similar shifts have been reported for different fixation treatments in other studies and could arise from a change in the optical refractive index mismatch within tissue[56-60] as well as chemical changes to the protein backbone. Moreover, the shift between the amide I and amide II bands may be the result of an altered microenvironment, loss of some chemical species, or exposure to atmosphere. The tissue that was sublimed with DHB but not subjected to MALDI MS analysis had the highest absorbance values, whereas the tissue analyzed by MALDI MS more closely resembled the fixed tissue section. The spectral differences may be from residual DHB absorbing in the region of

1500–1750 cm-1. Overall, most IR spectral features are similar between the adjacent tissue sections, demonstrating that performing MALDI MS before IR spectroscopy results only in minor changes to the IR imaging data. Compared to the drastic differences seen in MALDI MS performed after IR spectroscopy, the order of multimodal analysis described here provides an acceptable compromise.

Chemical Pan Sharpening of Ion Images.

A benefit of combining different imaging approaches is the increased flexibility in data analysis and visualization. Various types of microscopy data sets (e.g., haemotoxylin staining,

234 eosin staining, and electron microscopy) are utilized for pan sharpening of chemical imaging outputs. However, not all chemically unique morphological features can be observed in singleplex images. Also, using contrasting distributions can introduce artifacts or fail to improve sharpness, which is especially problematic when only a single image is available for pan sharpening. With additional choices for sharpening, pan sharpening can be optimized to enhance a given ion image without introducing artifacts. Multiplex IR datasets display a “mesoscale” level of detail in tissue composition, such as overall lipid, protein, and nucleotide distributions, but at higher spatial resolution than MSI. Thus, these images present unique opportunities for pan sharpening of MSI data sets. Through chemical pan sharpening, more ion images are suitable for sharpening as more corresponding contrast distributions can be found. Figures 7.7–7.12 show the effects of sharpening of single ion images with different IR bands. Two different lipid ion images (Figures 7.7–7.12C, F) are pan sharpened with absorbance images at two QCL-IR bands

(Figure 7.7–7.12A, B) associated with lipids and proteins. Overall, Figures 7.7, 7.9, 7.11D, E show similar distributions of signal intensities because Laplacian pyramid sharpening utilizes the low frequency components from the MS ion image. Within the DG (Figures 7.8, 7.10, 7.12), many small, cellular structures arise in sharpened images from high frequency information contained in the 1656 cm-1 band. However, because the cell-like structures are not seen in the ion image, the features must arise from IR absorbance contrast and cannot be used for making assumptions of analyte localization within the cell bodies. While more apparent in Figure S5E,

H, the data also demonstrate the ability of IR microscopy to detect fine tissue features that are invisible to MSI that may be native or arise during sample handling.

Careful handling is paramount during tissue sampling and sectioning as sample preparation artifacts can be propagated to the pan-sharpened image. In Figure 7.13, the

235 sharpened image in panel G appears blurrier than the image in panel H, which is sharpened by an

IR band exhibiting higher contrast around the DG. The example images demonstrate that the quality of pan sharpening depends on the contrast and distributions present in the underlying, higher spatial resolution images. Pan sharpening with IR distributions provides additional options for selecting appropriate bands for sharpening, but care is required when extrapolating beyond distributions present in the low spatial resolution, MS images.

Since Laplacian pyramid fusion functions on individual ion distributions, sharpened ion images can be combined and presented as an RGB composite (Figures 7.4, 7.13, 7.14). Figure

7.4A shows the RGB MS image overlaying the distribution of three compounds in the hippocampus. While different layers within the hippocampus are visible, the spatial resolution at which MSI was acquired is insufficient for distinguishing clear regional boundaries and, in some cases cellular localizations, within the brain tissue. However, sharpening each ion image with an absorbance image arising from the amide II band (1656 cm-1, Figures 7.4B, D and 7.13, 7.14, panels B, D) accentuates boundaries between different layers within the hippocampus. Focusing on the DG (Figures 7.4 and 7.13, 7.14, panels C, D) highlights the improvement in image quality, especially for replicates acquired at 50 µm×50 µm pixels (Figures 7.13, 7.14). After fusion, tissue morphology associated with known cellular structures becomes more apparent; e.g., the round structures within the bottom portion of the DG that appear as blurry, indistinct hot spots (Figure 7.4C). The resulting sharpened image is a qualitative fusion of the multimodal datasets, with color and brightness determined by ion abundances within a pixel and high frequency spatial information from the amide II stretching band.

Midlevel Data Fusion of IR and MALDI MS Images.

236

Hyperspectral PCA score images provide measures of the tissue variance, where pixels of the same color represent a closer distance in the score space and therefore similar chemical composition. Figure 7.5 shows hyperspectral images derived from spectral PCA of IR imaging

(Figure 7.5A) and FT-ICR-MSI (Figure 7.5B) data. The overlaid IR image produces a higher spatial resolution view of the DG with some cellular features visible (Figure 7.5A); however, only the first two principal components display meaningful morphological information as the third component appears to be noise. The hyperspectral PCA score images of FT-ICR MSI

(Figure 7.5B) have noticeably poorer spatial resolution but display more chemically distinct regions. Distributions of individual principal components higher than PC2 demonstrate more noise but depict some morphological detail (Figure 7.5B). Combining the PCA score images and performing PCA on the fused data produces the midlevel fusion image (Figure 7.5C). The fused dataset shows additional layers within the hippocampus at an improved spatial resolution. The improvement from midlevel fusion is especially striking for replicates shown in Figures 7.15,

7.16, where some artifacts visible in the IR images are removed from the fused images.

For biological replicate 1, acquired at 25 µm pixel widths for MSI, Davies-Bouldin optimization indicates the data is well modeled with 11 unique clusters (Figure 7.17), which associate well with histologically defined morphological regions within the hippocampus.

Notably, we can parse out the lucidum and alveus layers separately in segments 3 and 6, respectively. The unsupervised clustering effectively annotates the tissue morphology according to previous histological classifications. With a segment map, average IR and MS spectra for each region can reveal their unique chemical profiles. A representative example in shown in Figure

7.6 (with additional replicates shown in Figures 7.18–7.20); segments 1–4 are overlaid to show a clear chemical separation of CA1 (blue), the lucidum layer of the CA3 (pink), the rest of CA3

237

(green), and the DG (purple). A comparison of the average absorption spectra reveals differences around 1250cm-1, 1550 cm-1, and 1680 cm-1 (Figure 7.6B), which are largely related to differences in protein. Additionally, the average mass spectra show several, significant differences between phosphatidylcholine (PC) lipid signal intensities, as summarized in Figure

6C, D. The box plots (Figure 7.6D) show four example lipid species ([PC 32:0+K]+, [PC

34:1+K]+, [PC 36:4+K]+, and [PC 38:4+K]+) with differing abundance between segments corresponding to CA1, the lucidum layer of CA3, the rest of CA3, and the DG. Through data fusion and unsupervised clustering, significant chemical differences are detectable between physiologically similar regions. Neither method alone provides the chemical specificity and spatial resolution to differentiate between each layer. Multimodal imaging, optimized for high data quality, is vital for exploring chemical heterogeneity present in complex biological systems, including the brain.

7.4 Conclusions

We developed an analytical workflow that allows MALDI MS and IR spectroscopy to be performed on the same tissue. The workflow allows for chemical pan sharpening of MS ion images with a variety of IR absorption bands, providing optimized pan sharpened ion images with minimal artifacts. The inherent spatial agreement between modalities allowed us to determine lipid distributions within the DG, a morphologically and chemically complex, and functionally important, brain structure. Further, we developed a midlevel data fusion approach that allows chemical information from each technique to be combined, enhancing discrimination of structurally important regions by k-means clustering. Because MALDI MSI is a gentle ionization approach, laser damage was not observed in the IR images, making the incorporation of traditional staining methods feasible. Immunohistochemistry would further augment the

238 multimodal datasets with information on localization of targeted antigens within the hippocampus. The multimodal imaging approach presented here can be easily extended to other biologically important structures in the nervous system, such as the supraoptic and suprachiasmatic nuclei.

7.5 Acknowledgements

The authors gratefully acknowledge support from the National Institutes of Health,

Award Number 1U01 MH109062 from the National Institute of Mental Health and P30 018310 from the National Institute on Drug Abuse. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. E.K.N. acknowledges support from the National Science Foundation Graduate Research Fellowship

Program and the Springborn Fellowship. This work is also partially supported by the Agilent

Thought Leader Award to R.B.

239

7.6 Figures

Figure 7.1 Schematic of the combined MALDI-FT-ICR-MSI with IR imaging approach. A)

MALDI MSI is performed first, followed by (B) matrix removal and tissue fixation prior to (C)

IR imaging. Analyzing the same tissue sample simplifies (D) data fusion compared to the more common use of adjacent sections.

240

Figure 7.2 Influence of conducting IR spectroscopic imaging on the quality of MALDI MSI data. A) MALDI MS ion image of the hippocampus (m/z 798.5407; [PC 34:1 + K]+) for (i)

MALDI MS performed prior to IR spectroscopic imaging, (ii) after IR analysis with identical conditions – left side of brain section was analyzed, and (iii) right side of the same tissue section as in (ii) with twice the laser power and number of laser shots. B) Average, RMS normalized spectra for each image.

241

Figure 7.3 Influence of MALDI MSI on IR imaging data quality for rat hippocampus images. A)

Discrete frequency IR images of the 1248 cm-1 and 1656 cm-1 bands of adjacent tissue sections subjected to either (top) PFA treatment, (middle) sublimation of DHB with subsequent PFA treatment, or (bottom) sublimation of DHB and MALDI MSI analysis, with subsequent PFA treatment. B) The average absorbance spectra from the color-coded regions of the hippocampus are overlaid. While there are several morphological differences between the three tissue sections, the IR spectra match.

242

Figure 7.4 RGB composite ion image: R = [PC 32:0 + K]+ (m/z 772.5253), G = [PC 40:6 + K]+

(m/z 872.5566), and B = [PC 34:0 + K]+ (m/z 800.5567). A) Unsharpened MS image of the hippocampus taken with MSI at 25 µm spatial resolution. B) MS image in A, sharpened with

QCL-IR band at 1656 cm-1. C) Magnification of the area marked in A containing an unsharpened

MS image of the DG (magnified inset) and one end of the CA3 region. D) Magnification of sharpened MS image shown in B.

243

Figure 7.5 Hyperspectral PCA score images and individual principal components showing the expected structures within the DG region of the rat hippocampus. Images are from (A) IR spectra, (B) MALDI FT-ICR-MSI, and (C) midlevel data fusion of data presented in panels A and B.

244

Figure 7.6 Results of k-means clustering of fused data sets. A) Segments associated with the

CA1 (blue), lucidum layer of CA3 (pink), rest of CA3 (green), and the DG (purple) layers within the hippocampus. B) Average absorbance spectra for the four clusters. C) Average mass spectra for the specified clusters. Asterisks denote the m/z values used for the 4 boxplots presented in D.

D) Boxplots of selected signal intensities within each highlighted cluster. Labels above each box denote significantly different groups by post-hoc Tukey’s honestly significant difference multiple comparison test.

245

Figure 7.7 Biological replicate 1. Pan sharpening of mass spectrometry (MS) ion images by quantum cascade laser (QCL-IR) absorbance images. A) QCL-IR image of the hippocampus displaying absorbance at the ester band (1088 cm-1). B) QCL-IR image of the amide II band

(1656 cm-1). C) Matrix-assisted laser desorption/ionization-Fourier transform ion cyclotron resonance (MALDI-FT-ICR) MS ion image of [PC 40:6 + K]+ (m/z 772.5252). D, E) Ion image of [PC 40:6 + K]+ pan sharpened with the (D) ester band or the (E) amide II band. (F) MALDI-

FT-ICR MS ion image of [PC 34:0 + K]+ (m/z 872.5566). G, H) Ion images of [PC 34:0 + K]+ sharpened with the (G) ester band or the H) amide II band. MSI was acquired with 25 µm pixel spacing and QCL-IR was acquired with 5 µm pixel spacing.

246

Figure 7.8 Magnified images of Figure 7.7, focusing on the DG.

247

Figure 7.9 Biological replicate 2. In contrast to biological replicate 1, MSI was acquired at 50

µm pixel size. The boxed region in panel C corresponds to the magnified in region in Figure

7.10.

248

Figure 7.10 Magnified images of Figure 7.9, focusing on the DG.

249

Figure 7.11 Biological replicate 3 with MSI acquired at 50 µm pixel width. Boxed region in

Panel C corresponds to the DG region shown in Figure 7.12.

250

Figure 7.12 Magnified images of Figure 7.11, focusing on the DG.

251

Figure 7.13 Overlay of sharpened ion images for biological replicate 2 of Figure 7.4. The red, green and blue channels represent the ion intensities of [PC 32:0 + K]+ (772.5253), [PC 40:6 +

K]+ (872.5566), and [PC 34:0 + K]+ (800.5567), respectively. A) Interpolated MS image of the hippocampus taken at 50 µm spatial resolution. The boxed-in region is magnified in panels C and D. B) MS image sharpened with QCL-IR image acquired at 1656 cm-1. C) Interpolated MS image of hippocampal region containing one end of the CA3 region and the DG. D) Sharpened

MS image of the same area shown in C.

252

Figure 7.14 Overlay of sharpened ion images for biological replicate 3 of Figure 7.4. The red, green and blue channels represent the ion intensities of [PC 32:0 + K]+ (772.5253), [PC 40:6 +

K]+ (872.5566), and [PC 34:0 + K]+ (800.5567), respectively. A) Interpolated MS image of the hippocampus taken at 50 µm spatial resolution. The boxed-in region is shown in panels C and D.

B) MS image sharpened with the QCL-IR image acquired at 1656 cm-1. C) Interpolated MS image of hippocampal region containing one end of the CA3 region and the DG. D) Sharpened

MS image of the same area shown in C.

253

Figure 7.15 Hyperspectral score images of each modality and midlevel data fusion for biological replicate 2. FT-ICR-MSI data were acquired with 50 µm pixel widths. The overlay images (first row) are constructed by mapping principal components 1–3 to red, green and blue, respectively.

A) QCL-IR score image and corresponding single component images. B) FT-ICR-MSI score image with bilinear interpolation to align with the QCL-IR image area. C) Midlevel data fusion image of datasets shown in corresponding images in (A) and (B). Hippocampal region, containing one end of the CA3 region and the DG, is depicted.

254

Figure 7.16 Hyperspectral score images of each modality and midlevel data fusion for biological replicate 3. FT-ICR-MSI data were acquired with 50 µm pixel widths. The overlay images (first row) are constructed by mapping principal components 1–3 to red, green and blue respectively.

A) QCL-IR score image and corresponding single component images. B) FT-ICR-MSI score image with bilinear interpolation to align with QCL-IR image area. C) Midlevel data fusion image of data sets shown in corresponding images in (A) and (B). Hippocampal region, containing one end of the CA3 region and the DG, is depicted.

255

Figure 7.17 k-means clustering of the midlevel fused datasets of biological replicate 1, with k =

11. Each pixel in the image is grouped into a defined cluster, signified by a distinct color. k = 11 was found to be the optimum by Davies-Bouldin optimization metric.

256

Figure 7.18 k-means clustering of the (A) QCL-IR, (B) FT-ICR MS, and (C) midlevel fused datasets acquired from biological replicate 1, with k = 4. Each pixel in the image is grouped into a defined cluster, signified by a distinct color, shown individually in rows 2 through 5. Segments were manually assigned colors to match the highlighted morphology across modalities and biological replicates.

257

Figure 7.19 k-means clustering of the (A) QCL-IR, (B) FT-ICR MS, and (C) midlevel fused datasets of biological replicate 2, with k = 4. Each pixel in the image is grouped into a separate cluster, signified by a distinct color, shown individually in rows 2 through 5. Segments were manually assigned colors to match the highlighted morphology across modalities and biological replicates.

258

Figure 7.20 k-means clustering of the (A) QCL-IR, (B) FT-ICR MS, and (C) midlevel fused datasets of biological replicate 3, with k = 4. Each pixel in the image is grouped into a separate cluster, signified by a distinct color, shown individually in rows 2 through 5. Segments were manually assigned colors to match the highlighted morphology across modalities and biological

259

Figure 7.20 (cont.) replicates. Replicate 3 is a more caudal section of the brain than replicates 1 and 2, as identified by the absence of the granular layer adjacent to the midline (green in segment

2).

260

Figure 7.21 Boxplots of potassium and sodium adducts for the ions presented in the main text.

The intensity ratios between the salt adducts generally follow the same trend.

261

7.7 References

(1) Scoville, W. B.; Milner, B. Loss of recent memory after bilateral hippocampal lesions. J.

Neurol., Neurosurg. Psychiatry 1957, 20, 11-21.

(2) van Strien, N. M.; Cappaert, N. L. M.; Witter, M. P. The anatomy of memory: an interactive

overview of the parahippocampal–hippocampal network. Nat. Rev. Neurosci. 2009, 10,

272.

(3) Kee, N.; Teixeira, C. M.; Wang, A. H.; Frankland, P. W. Preferential incorporation of adult-

generated granule cells into spatial memory networks in the dentate gyrus. Nat. Neurosci.

2007, 10, 355.

(4) Bruel-Jungerman, E.; Laroche, S.; Rampon, C. New neurons in the dentate gyrus are

involved in the expression of enhanced long-term memory following environmental

enrichment. Eur. J. Neurosci. 2005, 21, 513-521.

(5) Seri, B.; García-Verdugo, J. M.; Collado-Morente, L.; McEwen, B. S.; Alvarez-Buylla, A.

Cell types, lineage, and architecture of the germinal zone in the adult dentate gyrus. J.

Comp. Neurol. 2004, 478, 359-378.

(6) James Surmeier, D.; Wilson, C. J.; Eberwine, J. Patch-Clamp Techniques for Studying

Potassium Currents in Mammalian Brain Neurons, 1994; Vol. 19, p 39-67.

(7) Jennes, L.; Stumpf, W. E.; Kalivas, P. W. Neurotensin: Topographical distribution in rat

brain by immunohistochemistry. J. Comp. Neurol. 1982, 210, 211-224.

(8) Glover LR, S. T., Karlsson R-M, Bannerman DM, Cameron HA Ongoing neurogenesis in the

adult dentate gyrus mediates behavioral responses to ambiguous threat cues. PLoS Biol.

2017, 15, e2001154.

262

(9) Jones, D. K. Studying connections in the living human brain with diffusion MRI. Cortex

2008, 44, 936-952.

(10) Fenno, L.; Yizhar, O.; Deisseroth, K. The Development and Application of Optogenetics.

Annu. Rev. Neurosci. 2011, 34, 389-412.

(11) Livet, J.; Weissman, T. A.; Kang, H.; Draft, R. W.; Lu, J.; Bennis, R. A.; Sanes, J. R.;

Lichtman, J. W. Transgenic strategies for combinatorial expression of fluorescent

proteins in the nervous system. Nature 2007, 450, 56.

(12) Heidenreich, M.; Zhang, F. Applications of CRISPR–Cas systems in neuroscience. Nat.

Rev. Neurosci. 2015, 17, 36.

(13) Okaty, B. W.; Sugino, K.; Nelson, S. B. Cell Type-Specific Transcriptomics in the Brain. J.

Neurosci. 2011, 31, 6939-6943.

(14) El-Aneed, A.; Cohen, A.; Banoub, J. Mass Spectrometry, Review of the Basics:

Electrospray, MALDI, and Commonly Used Mass Analyzers. Appl. Spectrosc. Rev.

2009, 44, 210-230.

(15) Gessel, M. M.; Norris, J. L.; Caprioli, R. M. MALDI imaging mass spectrometry: Spatial

molecular analysis to enable a new age of discovery. Journal of Proteomics 2014, 107,

71-82.

(16) Zavalin, A.; Todd, E. M.; Rawhouser, P. D.; Yang, J.; Norris, J. L.; Caprioli, R. M. Direct

imaging of single cells and tissue at sub-cellular spatial resolution using transmission

geometry MALDI MS. J. Mass Spectrom. 2012, 47, 1473-1481.

(17) Kompauer, M.; Heiles, S.; Spengler, B. Atmospheric pressure MALDI mass spectrometry

imaging of tissues and cells at 1.4-μm lateral resolution. Nat. Methods 2016, 14, 90.

263

(18) Jurchen, J. C.; Rubakhin, S. S.; Sweedler, J. V. MALDI-MS imaging of features smaller

than the size of the laser beam. J. Am. Soc. Mass Spectrom. 2005, 16, 1654-1659.

(19) Van de Plas, R.; Yang, J.; Spraggins, J.; Caprioli, R. M. Image fusion of mass spectrometry

and microscopy: a multimodality paradigm for molecular tissue mapping. Nat. Methods

2015, 12, 366.

(20) Tarolli, J. G.; Jackson, L. M.; Winograd, N. Improving Secondary Ion Mass Spectrometry

Image Quality with Image Fusion. J. Am. Soc. Mass Spectrom. 2014, 25, 2154-2162.

(21) Vollnhals, F.; Audinot, J.-N.; Wirtz, T.; Mercier-Bonin, M.; Fourquaux, I.; Schroeppel, B.;

Kraushaar, U.; Lev-Ram, V.; Ellisman, M. H.; Eswara, S. Correlative Microscopy

Combining Secondary Ion Mass Spectrometry and Electron Microscopy: Comparison of

Intensity–Hue–Saturation and Laplacian Pyramid Methods for Image Fusion. Anal.

Chem. 2017, 89, 10702-10710.

(22) Bhargava, R. Infrared Spectroscopic Imaging: The Next Generation. Appl. Spectrosc. 2012,

66, 1091-1120.

(23) Nasse, M. J.; Walsh, M. J.; Mattson, E. C.; Reininger, R.; Kajdacsy-Balla, A.; Macias, V.;

Bhargava, R.; Hirschmugl, C. J. High-resolution Fourier-transform infrared chemical

imaging with multiple synchrotron beams. Nat. Methods 2011, 8, 413.

(24) Fernandez, D. C.; Bhargava, R.; Hewitt, S. M.; Levin, I. W. Infrared spectroscopic imaging

for histopathologic recognition. Nat. Biotechnol. 2005, 23, 469.

(25) Hackett, M. J.; McQuillan, J. A.; El-Assaad, F.; Aitken, J. B.; Levina, A.; Cohen, D. D.;

Siegele, R.; Carter, E. A.; Grau, G. E.; Hunt, N. H. Chemical alterations to murine brain

tissue induced by formalin fixation: implications for biospectroscopic imaging and

mapping studies of disease pathogenesis. Analyst 2011, 136, 2941-2952.

264

(26) Bergner, N.; Romeike, B. F.; Reichart, R.; Kalff, R.; Krafft, C.; Popp, J. Tumor margin

identification and prediction of the primary tumor from brain metastases using FTIR

imaging and support vector machines. Analyst 2013, 138, 3983-3990.

(27) Kastyak-Ibrahim, M.; Nasse, M.; Rak, M.; Hirschmugl, C.; Del Bigio, M.; Albensi, B.;

Gough, K. M. Biochemical label-free tissue imaging with subcellular-resolution

synchrotron FTIR with focal plane array detector. NeuroImage 2012, 60, 376-383.

(28) Leslie, L. S.; Wrobel, T. P.; Mayerich, D.; Bindra, S.; Emmadi, R.; Bhargava, R. High

definition infrared spectroscopic imaging for lymph node histopathology. PLoS ONE

2015, 10, e0127238.

(29) Mayerich, D.; Walsh, M. J.; Kadjacsy-Balla, A.; Ray, P. S.; Hewitt, S. M.; Bhargava, R.

Stain-less staining for computed histopathology. Technology 2015, 3, 27-31.

(30) Bhargava, R.; Madabhushi, A. Emerging themes in image informatics and molecular

analysis for digital pathology. Annu. Rev. Biomed. Eng. 2016, 18, 387-412.

(31) Baker, M. J.; Trevisan, J.; Bassan, P.; Bhargava, R.; Butler, H. J.; Dorling, K. M.; Fielden,

P. R.; Fogarty, S. W.; Fullwood, N. J.; Heys, K. A. Using Fourier transform IR

spectroscopy to analyze biological materials. Nat. Protoc. 2014, 9, 1771.

(32) Masyuko, R.; Lanni, E. J.; Sweedler, J. V.; Bohn, P. W. Correlated imaging - a grand

challenge in chemical analysis. Analyst 2013, 138, 1924-1939.

(33) Lanni, E. J.; Masyuko, R. N.; Driscoll, C. M.; Dunham, S. J. B.; Shrout, J. D.; Bohn, P. W.;

Sweedler, J. V. Correlated Imaging with C60-SIMS and Confocal Raman Microscopy:

Visualization of Cell-Scale Molecular Distributions in Bacterial Biofilms. Anal. Chem.

2014, 86, 10885-10891.

265

(34) Li, Z.; Chu, L.-Q.; Sweedler, J. V.; Bohn, P. W. Spatial Correlation of Confocal Raman

Scattering and Secondary Ion Mass Spectrometric Molecular Images of Lignocellulosic

Materials. Anal. Chem. 2010, 82, 2608-2611.

(35) Petit, V. W.; Réfrégiers, M.; Guettier, C.; Jamme, F.; Sebanayakam, K.; Brunelle, A.;

Laprévote, O.; Dumas, P.; Le Naour, F. Multimodal Spectroscopy Combining Time-of-

Flight-Secondary Ion Mass Spectrometry, Synchrotron-FT-IR, and Synchrotron-UV

Microspectroscopies on the Same Tissue Section. Anal. Chem. 2010, 82, 3963-3968.

(36) Le Naour, F.; Bralet, M.-P.; Debois, D.; Sandt, C.; Guettier, C.; Dumas, P.; Brunelle, A.;

Laprévote, O. Chemical Imaging on Liver Steatosis Using Synchrotron Infrared and ToF-

SIMS Microspectroscopies. PLoS ONE 2009, 4, e7408.

(37) Lasch, P.; Noda, I. Two-Dimensional Correlation Spectroscopy for Multimodal Analysis of

FT-IR, Raman, and MALDI-TOF MS Hyperspectral Images with Hamster Brain Tissue.

Anal. Chem. 2017, 89, 5008-5016.

(38) Page, J. S.; Sweedler, J. V. Sample Depletion of the Matrix-Assisted Laser Desorption

Process Monitored Using Radionuclide Detection. Anal. Chem. 2002, 74, 6200-6204.

(39) Comi, T. J.; Neumann, E. K.; Do, T. D.; Sweedler, J. V. microMS: A Python Platform for

Image-Guided Mass Spectrometry Profiling. J. Am. Soc. Mass Spectrom. 2017, 28, 1919-

1928.

(40) Comi, T. J.; Makurath, M. A.; Philip, M. C.; Rubakhin, S. S.; Sweedler, J. V. MALDI MS

Guided Liquid Microjunction Extraction for Capillary Electrophoresis–Electrospray

Ionization MS Analysis of Single Pancreatic Islet Cells. Anal. Chem. 2017, 89, 7765-

7772.

266

(41) Dufresne, M.; Guneysu, D.; Patterson, N. H.; Marcinkiewicz, M. M.; Regina, A.; Demeule,

M.; Chaurand, P. Multimodal detection of GM2 and GM3 lipid species in the brain of

mucopolysaccharidosis type II mouse by serial imaging mass spectrometry and

immunohistochemistry. Anal. Bioanal. Chem. 2017, 409, 1425-1433.

(42) Kaya, I.; Michno, W.; Brinet, D.; Iacone, Y.; Zanni, G.; Blennow, K.; Zetterberg, H.;

Hanrieder, J. Histology-Compatible MALDI Mass Spectrometry Based Imaging of

Neuronal Lipids for Subsequent Immunofluorescent Staining. Anal. Chem. 2017, 89,

4685-4694.

(43) Chaurand, P.; Schwartz, S. A.; Billheimer, D.; Xu, B. J.; Crecelius, A.; Caprioli, R. M.

Integrating Histology and Imaging Mass Spectrometry. Anal. Chem. 2004, 76, 1145-

1155.

(44) Bich, C.; Vianello, S.; Guérineau, V.; Touboul, D.; De La Porte, S.; Brunelle, A.

Compatibility between TOF-SIMS lipid imaging and histological staining on a rat brain

section. Surf. Interface Anal. 2013, 45, 260-263.

(45) Deutskens, F.; Yang, J.; Caprioli, R. M. High spatial resolution imaging mass spectrometry

and classical histology on a single tissue section. J. Mass Spectrom. 2011, 46, 568-571.

(46) Brand-Williams, W.; Cuvelier, M. E.; Berset, C. Use of a free radical method to evaluate

antioxidant activity. LWT--Food Sci. Technol. 1995, 28, 25-30.

(47) Ashidate, K.; Kawamura, M.; Mimura, D.; Tohda, H.; Miyazaki, S.; Teramoto, T.;

Yamamoto, Y.; Hirata, Y. Gentisic acid, an aspirin metabolite, inhibits oxidation of low-

density lipoprotein and the formation of cholesterol ester hydroperoxides in human

plasma. Eur. J. Pharmacol. 2005, 513, 173-179.

267

(48) Sproß, J. Analysis of Volatile and Oxidation Sensitive Compounds Using a Cold Inlet

System and Electron Impact Mass Spectrometry. J.Vis. Exp. 2014, 51858.

(49) Aparicio, S.; Doty, S.; Camacho, N.; Paschalis, E.; Spevak, L.; Mendelsohn, R.; Boskey, A.

Optimal methods for processing mineralized tissues for Fourier transform infrared

microspectroscopy. Calcif. Tissue Int. 2002, 70, 422-429.

(50) Lyng, F.; Gazi, E.; Gardner, P. In Biomedical Applications of Synchrotron Infrared

Microspectroscopy: A Practical Approach Moss, D., Ed.; Royal Society of Chemistry:

Cambridge, 2011, pp 147-185.

(51) Kwak, J. T.; Reddy, R.; Sinha, S.; Bhargava, R. Analysis of variance in spectroscopic

imaging data from human tissues. Anal. Chem. 2011, 84, 1063-1069.

(52) Lanni, E. J.; Dunham, S. J. B.; Nemes, P.; Rubakhin, S. S.; Sweedler, J. V. Biomolecular

Imaging with a C60-SIMS/MALDI Dual Ion Source Hybrid Mass Spectrometer:

Instrumentation, Matrix Enhancement, and Single Cell Analysis. J. Am. Soc. Mass

Spectrom. 2014, 25, 1897-1907.

(53) Robichaud, G.; Garrard, K. P.; Barry, J. A.; Muddiman, D. C. MSiReader: An Open-Source

Interface to View and Analyze High Resolving Power MS Imaging Files on Matlab

Platform. J. Am. Soc. Mass Spectrom. 2013, 24, 718-721.

(54) Burt, P.; Adelson, E. The Laplacian Pyramid as a Compact Image Code. IEEE Transactions

on Communications 1983, 31, 532-540.

(55) Garrett, T. J.; Prieto-Conaway, M. C.; Kovtoun, V.; Bui, H.; Izgarian, N.; Stafford, G.; Yost,

R. A. Imaging of small molecules in tissue sections with a new intermediate-pressure

MALDI linear ion trap mass spectrometer. Int. J. Mass spectrom. 2007, 260, 166-176.

268

(56) Mohlenhoff, B.; Romeo, M.; Diem, M.; Wood, B. R. Mie-type scattering and non-Beer-

Lambert absorption behavior of human cells in infrared microspectroscopy. Biophys. J.

2005, 88, 3635-3640.

(57) Bassan, P.; Byrne, H. J.; Bonnier, F.; Lee, J.; Dumas, P.; Gardner, P. Resonant Mie

scattering in infrared spectroscopy of biological materials–understanding the ‘dispersion

artefact’. Analyst 2009, 134, 1586-1593.

(58) Perez-Guaita, D.; Heraud, P.; Marzec, K. M.; De La Guardia, M.; Kiupel, M.; Wood, B. R.

Comparison of transflection and transmission FTIR imaging measurements performed on

differentially fixed tissue sections. Analyst 2015, 140, 2376-2382.

(59) Davis, B. J.; Carney, P. S.; Bhargava, R. Theory of Mid-infrared Absorption

Microspectroscopy: II. Heterogeneous Samples. Anal. Chem. 2010, 82, 3487-3499.

(60) Davis, B. J.; Carney, P. S.; Bhargava, R. Theory of Midinfrared Absorption

Microspectroscopy: I. Homogeneous Samples. Anal. Chem. 2010, 82, 3474-3486.

269

CHAPTER 8

SINGLE CELL MALDI MASS SPECTROMETRY HYPHENATED TO STIMULATED RAMAN

SCATTERING MICROSCOPY FOR ENHANCED CHEMICAL COVERAGE

Notes and Acknowledgements

This chapter is adapted from a completed manuscript coauthored by Troy J. Comi,

Sangamitra Deb, Stanislav S. Rubakhin, Martha U. Gillette, Rohit Bhargava, and Jonathan V.

Sweedler, submitted for publication in 2019. T.J. Comi performed some data analysis and wrote all scripts used in the study. S. Deb performed the SRSM analysis. S.S. Rubakhin, M.U. Gillette,

R. Bhargava, and J.V. Sweedler assisted with concept development and writing, as well as funded the work. We acknowledge Dr. Jennifer Mitchell for useful discussions. The authors gratefully acknowledge support from the National Institutes of Health, Award Number P30

DA018310 from the National Institute on Drug Abuse, and from the National Institute of Mental

Health, Award Number 1U01 MH109062. E.K. Neumann and T.J. Comi acknowledge support from the National Science Foundation Graduate Research Fellowship Program and the

Springborn Fellowship. T.J. Comi received additional support through the Training Program at

Chemistry-Interface with Biology (T32 GM070421). This work is also partially supported by the

Agilent Thought Leader Award to R. Bhargava.

8.1 Introduction

The brain is among the most structurally complex and chemically heterogeneous biological systems, in which cell-to-cell interactions regulate a variety of essential physiological functions such as learning and memory. Thus, single cell resolution measurements are vital for studying complex tissues of this type. Single cell analyses are possible with a variety of analytical methods and the capability to perform meaningful measurements at the single cell

270 level are an important benchmark of sensitivity and detectability. Small sample volumes, low absolute chemical abundances, and difficulties with sample handling limit the chemical coverage of most direct single cell analytical measurements. Extending these limits remains an area of active research.

Genomic and transcriptomic approaches increase analytical sensitivity using amplification to bring concentration into the range of sensitive measures. Single cell transcriptomics has dominated single cell measurements in recent years[1-2] and studies on neurons[3] and glia[4] have revealed genetic heterogeneity with functional consequences, underscoring the need for greater understanding at the single cell level. In particular, greatly complementing transcriptomics, the direct chemical analysis of the peptide and metabolic contents of single cells provides a close link to cellular phenotypes.[5]

Mass spectrometry (MS) is a powerful method for single cell measurements with high sensitivity and molecular specificity. We have developed microscopy-guided single cell matrix assisted laser desorption/ionization (MALDI) MS and applied this approach to a variety of rodent cellular populations.[6-9] Microscopy-guided MALDI MS analysis rates approach 1 Hz facilitating measurements of thousands of cells for tens to hundreds of chemicals, including lipids, peptides, and small metabolites. While efforts to reduce the MALDI laser spot size[10-11] or utilize oversampling methods have driven subcellular MS imaging,[12-13] such measurements are still restricted to specialty labs; commercial instrumentation more commonly achieves >10 µm spatial resolutions. Microscopy-guided profiling circumvents limitations in the spatial resolution of MALDI lasers by analyzing dispersed cells, however, subcellular morphology and original location within the tissue is typically lost.

271

Vibrational spectroscopy is a non-destructive method that can rapidly obtain molecular information and techniques such as stimulated Raman scattering microscopy (SRSM) now offer submicron (<0.5 µm) spatial resolution.[14-15] Further, vibrational spectroscopy is sensitive to changes in pH, depth, analyte concentrations, mixture complexity, and temperature, providing information beyond chemical composition. Unfortunately, SRSM spectral resolutions alone cannot identify chemical species in complex mixtures and instead requires standards, labels, or orthogonal approaches to verify chemical identities beyond class.[16] While multi-wavelength cytotyping with infrared vibrational spectroscopic imaging is feasible at micron-scale pixels, these spectra serve as a barcode and cannot identify specific molecules.

Multimodal chemical analyses are becoming more common for investigations of complex chemical systems. Ideally, interrogating a sample with two complementary approaches provides chemical information greater than either approach alone.[17-20] Raman spectroscopy and MS have been performed on the same tissue section[21-25] often at near-cellular resolutions[26] as well as single cell samples.[27-28] Pioneering work by the Zenobi group correlated fluorescence microscopy, MS, and confocal Raman microscopy for single algal cells using microarrays for

MS to study lower molecular weight metabolites.[27-28]

To extend single cell multimodal analysis, we combine SRSM with MALDI FT-ICR MS to relate the molecular specificity of MS with subcellular images from SRSM for several hundred rodent hippocampal cells. Although SRSM is non-destructive, we found performing

MALDI MS first to be optimal as it improves the effective throughput of SRSM acquisition by targeting only cells with adequate MS signal intensity. By combining the two techniques on the same single cell, we can achieve 250 nm resolution images of broad chemical classes (e.g. lipids, nucleic acids, and proteins) at higher sensitivity using SRSM as opposed to more traditional

272

Raman spectroscopic approaches, while still detecting discrete phospholipids, such as phosphatidylcholine (PC)(32:0), PC(34:1), PC(36:1), and phosphatidic acid (PA)(38:4) with the confidence afforded by FT-ICR MS measurements.

8.2 Experimental

Chemicals.

All chemicals were purchased from Sigma Aldrich and used without further purification unless otherwise specified.

Animals.

2–2.5 month-old male Sprague Dawley outbred rats (Rattus norvegicus)

(www.envigo.com) were housed on a 12 h light cycle and fed ad libitum. Animal euthanasia was performed in accordance with the appropriate institutional animal care guidelines (the

Illinois Institutional Animal Care and Use Committee), and in full compliance with federal guidelines for the humane care and treatment of animals. Three rats were used in these experiments.

Cell Dissociation.

Dissected hippocampal tissues were treated with papain dissociation system

(Worthington Biochemical, Lakewood, NJ). Briefly, tissues were incubated with an oxygenated solution of 20 units/mL of papain, 1 mM L-cysteine, and 0.5 mM ethylenediaminetetraacetic acid dissolved in Earle's balanced salt solution for 80 minutes at 34 °C. Tissue was then mechanically dissociated in modified Gey’s balanced salt solution (mGBSS) containing (in mM): 1.5 CaCl2, 5 KCl, 0.2 KH2PO4, 11 MgCl2, 0.3 MgSO4, 138 NaCl, 28 NaHCO3, and 0.8

Na2HPO4, and 25 HEPES, pH 7.2 and supplemented with 0.04% paraformaldehyde (PFA) for cell stabilization through mechanical dissociation. Following dissociation, glycerol was added to

273 a final concentration of 40% (v/v) and 50 µL of cell suspension was transferred onto non-coated microscopy glass slides (LabMed, Ft. Lauderdale, Florida) that had been etched with fiducial marks and washed with ethanol prior to plating. Although conductive glass slides are more common for MS analysis, they are incompatible with SRSM analysis. Importantly, MALDI FT-

ICR MS can analyze samples deposited on non-conductive substrates.

The glass slide held deposits of cells from three animals at separate locations. After ~18 h, cells were stained with Hoechst 33342 (1 µg/mL in mGBSS) for 15 min before rinsing twice with 500 µL of 150 mM ammonium acetate and drying under a slow stream of nitrogen gas.

Optical Imaging.

Brightfield and fluorescence images were acquired on a Zeiss Axio M2 microscope

(Zeiss, Jena, Germany) equipped with an Ab cam Icc5 camera, X-cite Series 120 Q mercury lamp (Lumen Dynamics, Mississauga, Canada), and a HAL 100 halogen illuminator (Zeiss, Jena,

Germany). The DAPI dichroic (ex. 335-383 nm; em. 420-470 nm) was used for fluorescence excitation. The images were acquired with a 10x objective (1-pixel width is 0.55 µm) with a 13% overlap produced during image tiling. Images were processed and exported as big tiff files on

ZEN 2 software (blue edition, Zeiss, Jena, Germany).

Matrix Application and MS Analysis.

2,5-dihydroxybenzoic acid (DHB) was sublimed onto single cell deposits forming a fine crystal layer with a thickness of ~0.3 mg/cm2 using a custom sublimation chamber described previously.17 Following sublimation, slides returned to 24 °C within a dry nitrogen box for at least 15 min.

Single cell MALDI mass spectra were acquired on a solariX XR 7T FT-ICR mass spectrometer (Bruker Corp., Billerica, MA) with a mass window of m/z 150-3000, yielding a

274 transient length of 2.94 sec. The “ultra” laser setting was used with a laser footprint of approximately 100 µm. 50 laser shots were accumulated with a frequency of 1000 Hz at 50% of the maximum power settings. Single cell coordinates and a custom geometry file were ascertained from microMS, as previously described,[8] as well as an Excel file for the target automation function on ftmsControl (v. 2.1.0, Bruker Corp.). Cells were distance filtered to be

100 µm apart and filtered by size to remove fluorescent debris. Mass spectrum acquisition order was randomized across all three animals on the slide to limit any time-dependent batch effects.

MALDI FT-ICR mass spectra were displayed in Data Analysis (v4.4, Bruker Corp.) with a linear recalibration using [phosphatidylcholine (PC)(32:0)+H]+, [PC(32:0)+Na]+, and [PC(38:4)+H]+.

Matrix Removal and Sample Fixation.

After MS analysis, hippocampal cell dissociates were placed within a solution of 4% PFA in 1x phosphate buffered saline (PBS; ThermoFisher Scientific, Waltham, MA) for 10 min. The

PFA in PBS solution was replaced with fresh solution and incubation continued for an additional

5 min to remove MALDI matrix with concurrent cell fixation. Samples were washed in PBS for

3 min following fixation and briefly rinsed with Mili-Q water to remove remaining salts. Finally, slides were dried under a stream of nitrogen gas and stored in a vacuum desiccator until SRSM analysis.

SRSM Analysis.

The microMS software was adapted to allow image registration with a custom SRSM system. The SRSM coordinate space was registered with optical images by extending microMS as previously reported.8 The stage coordinates of each cell were then used to direct acquisition of a ~50 µm2 image. SRSM control was slightly modified to acquire cell microarray images.

275

SRS chemical images were acquired using a two-photon laser scanning microscopy system built in house.[34] The SRS imaging setup is integrated with a dual-output (1064 nm/532 nm, 80 MHz) ultrafast oscillator (Lumera, Coherent Germany) which pumps an optical parametric oscillator (OPO) (Levante Emerald, APE Germany) to provide tunable (750 nm – 970 nm) ~6 ps pulse trains of light. The OPO output is used as the pump pulse which is tuned to a

Raman mode of interest, while the 1064 nm beam is used as the Stokes pulse. The beams are spatiotemporally overlapped and sent collinearly to the SRS microscope. The Stokes pulse train is amplitude-modulated at 7 MHz by an electro-optic modulator (EOM, Conoptics). The images were acquired using a 50X (0.95 NA, Zeiss) objective and a home-built large area photodiode

(PS100-6, first sensor) detector. A high OD band-pass filter transmits the pump beam and blocks the Stokes beam for stimulated Raman loss (SRL) detection.[35] The detector has two amplification stages and separates the pump signal into AC and DC components. The SRL signals (based on transferred modulation from the Stokes beam to the pump beam) present in the

AC component are demodulated and amplified by a lock-in amplifier (HF2LI, Zurich

Instruments) with a time constant of 5 µs. Images were acquired with a 20 μs dwell time. The pump wavelength was tuned from 801 nm to 821 nm with 0.3 nm spacing to generate stimulated

Raman images. The selected cells were imaged from 2780 cm-1 to 3080 cm-1 with 4.5 cm-1 spacing (68 spectral points). Intensities at each frequency were power normalized during image analysis with respect to pump pulse power at the given pump wavelength. Total laser power at the sample plane was < 30 mW.

Data Registration and Analysis.

Since both MALDI MS and SRSM utilized the same optical image and cell positions, data coregistration was simplified to matching the optical image x,y coordinates between the

276 analyses. SRSM datasets were processed in MATLAB (The Mathworks, Nantucket, MA). The cells were imaged together to minimize the power variation for a given frequency. Cell images were normalized to pump power, accounting for power fluctuation during pump laser tuning.

Cell boundaries and average spectra were determined with a scheme based on principal component analysis (PCA) outlier detection. The first principal component scores were estimated from a random subset of 105 pixels. From the log transformed the first principal component values, a t-distributed, 99% confidence interval was determined. Pixels within the interval were non-cellular background signal and removed, while the remaining 1% corresponded to cellular material. The binary image produced from the first principal component from each cell was dilated and eroded with a 3 by 3 square to fill in single pixel gaps, where the largest area was concerned a cell. SRS spectra within the mask were averaged, smoothed, max normalized, and background corrected.

8.3 Results and Discussion

Procedure for multimodal analysis.

There are several considerations for single cell multi-modal spectroscopic imaging.

Single cell locations were determined using whole-slide fluorescent microscopy of Hoechst

33342 as previously described[8] (Figure 8.1A). In total, 9,276 cell-like structures could be found using microMS. After filtering the structures to maintain a minimum 100 µm nearest neighbor distance (the size of the laser footprint) and by size to reduce the likelihood of analyzing debris, only 1,342 cells remained for single cell MALDI FT-ICR MS analysis (Figure 8.1B). While microMS has an option to set a minimum size for cell finding, debris can also contaminate single cell spectra and performing the size filter afterwards reduces contamination from neighboring objects. A second consideration at this stage is the acquisition time for each of the multiple

277 modalities. Using the Bruker solariX target automation routines, we acquired a single cell spectrum every 30 sec. SRSM takes approximately 10 min to acquire a single cell chemical image for the full CH spectral range (including stage movement, calibration, and alignment), so performing MALDI MS prior to vibrational spectroscopy allows us to prescreen cells for SRSM analysis, significantly reducing SRSM sample characterization time. Third, we considered the effect of characterization order. Performing MS prior to vibrational spectroscopy preserves high- quality mass spectra of the cells, similarly to that of tissue sections.[36] Finally, our choice of techniques is dictated by the workflow. We use MALDI FT-ICR MS, because it allows analysis of cells placed on nonconductive sample substrates that are compatible with SRSM measurements without large mass shifts and lowered sensitivity. Additionally, tandem MS is difficult to perform on single cells, especially after SRSM analysis (cells are fixed) and for rare molecules, so the extra confidence afforded by high mass accuracy FT-ICR MS measurements is imperative for confident chemical assignments.

Following washing and fixation, SRSM was only performed on cells with adequate lipid signal from MALDI MS (596 cells; Figure 8.1C). Putative cells with insufficient lipid signatures likely correspond to cells damaged during the tissue dissociation process, or debris produced by sampling. While microMS has been used in previous reports for single cell MS analysis, here we adapted the software to direct the SRSM acquisition to targeted cells. microMS allows SRSM to image a 50 by 50 µm region encompassing each cell. Collecting Raman chemical images around each cell saves a significant amount of time over a single image, because most of the sample surface is empty. While usually not a concern in SRSM imaging experiments, distance between cells is a requirement to prevent cell-to-cell contamination during single cell MS analysis.

Because the same microscope image and software is used for both MALDI MS and SRSM

278 analysis, correlating the mass spectra with the vibration spectra is trivial (Figure 8.1D) as the same pixel coordinate is used in both instances.

Single cell spectra.

Using the developed method, mass spectra and high spatial resolution (250 nm) SRSM images are acquired from 596 cells isolated from three different rats. For this initial report, we did not perform detailed analysis of all analyzed cells, but instead highlight the information that can be garnered from hyphenated MALDI-MS and SRSM by choosing representative cells from each animal. A cell from each animal was chosen to demonstrate the performance of the method

(Figure 8.2). While the cells possess similar lipid contents, the relative intensities of corresponding signals differ, suggesting the need for discriminating individual lipid species at cellular resolutions. Similarly, there are several lipids only present within one cell, despite many being contained in all three (Table 8.1). The diversity and importance of lipids in the brain is well known,37-40 although single cell measurements are rarely used in accessing cellular lipid heterogeneity.41-42 Because the MALDI laser probe entirely covers the targeted cell, we use MS to collect the average lipid profile of each cell, disregarding subcellular lipid distributions, but obtaining specific lipid identities.

Overall, hundreds of spectral features with a signal-to-noise ratio above 3 were detected within each cell, demonstrating accessible chemical information contained within single cell MS data. For the cells featured here, 36, 37, and 27 discrete lipid features were detected within the cells from animals 1, 2, and 3, respectively. Most of the lipids were phospholipids, usually containing a PC headgroup (Figure 8.2.i), and all but 11 could be assigned using common databases (Table 8.1). While MS analysis here primarily detects phospholipids, MALDI MS can be used to study a variety of different chemical classes[43] by altering sample preparation, many

279 of which are compatible with our developed method. Additionally, the higher confidence afforded by FT-ICR MS as compared to time of flight-MS is important for these single cell, multimodal experiments, as the lipids are removed during the fixation step prior to SRSM analysis. Performing tandem MS between the MS and SRS analyses on single cells is difficult and would perhaps result in cellular damage, reducing the quality of SRS images.

The high spatial resolution SRSM images reveal subcellular morphology and provide chemical maps of cell contents (Figure 8.2.iii). Generally, signals for a higher abundance of lipids (red, 2845 cm-1) are seen along the outer edges of the cells with more protein and nucleic acid signatures within the cellular nucleus region (blue, 2930 - 2845 cm-1).[44] These frequencies are considered to correspond to lipids and a combination of proteins and nucleic acids, respectively, although we emphasize that it is not possible to unambiguously assign every voxel in a complex cellular milieu to these simple categories. Nevertheless, a limited set of frequencies provides sufficient contrast allowing identification of cellular material. Furthermore, the high spatial resolution images allow us to determine the average SRS spectra for nuclei compared to cell body (Figure 8.2.ii). We used presence of a nuclei, size, and MS signal to determine if an object was a cell or piece of debris.

Interestingly, MS lipid intensities do not always correlate with SRS lipid signal. We believe that the discrepancy between the two datasets results from many factors, including

MALDI insufficiently ionizing all classes of lipids equally and the SRS signal of non-lipid compounds also at 2845 cm-1. Although a direct correlation of the two approaches is not made for lipids, it demonstrates that MALDI MS and SRSM are sensitive to different, although overlapping, chemical constituents within the cell. By taking advantage of the orthogonal

280 chemical information that is detected by multiple approaches, the combined data can further discriminate or classify biological systems than either technique alone.[22, 29]

Beyond chemical information, the multimodal approach allows us to better assess microscopy-guided MS. SRSM was used to image 50 by 50 µm2 areas with the target cell at the center of each area and, as most images contain the target cell, we conclude microMS accurately correlates the image location of a cell with the SRSM stage position. SRSM images which do not contain a cell are most likely caused by the cell being removed during washing, though previous work indicates cell dislocation is a rare event (data not shown). Additionally, sample from animal 2 (Figure 8.3) shows two distinct cells in the SRSM image, as apparent from the nuclei and lipid signal between. The two cells were counted as a single feature in microMS because the nuclei overlap and are not distinguished during cell finding. Therefore, higher spatial resolution

SRSM imaging provides additional opportunity for single cell data verification. Moreover, the mild sample treatment is exemplified by the close concordance between the brightfield image, acquired prior to matrix coating, and SRSM images. Fine cell structures including membrane shape survive the analysis process, demonstrating a minimal disturbance of cellular structure that enables both sets of chemically-rich approaches to be performed on the same single cell.

8.4 Conclusions

We have successfully demonstrated hyphenating MALDI FT-ICR MS and SRSM imaging for the same cell. The combined data allows MS characterization of discrete phospholipids within the cells without sacrificing high spatial resolution chemical maps provided by vibrational spectroscopy. Further work is required for method optimization to reduce batch variances and allow for more confident population analyses. Moreover, 3D chemical images or live cell images can be acquired with SRSM, allowing the correlated data from MALDI MS-

281

SRSM to be applied to other systems where MALDI cannot be performed. Extending the methodology to other biological systems will allow the exploration of functionally important chemicals within dynamic physiological processes, such as memory or circadian rhythms.

Analyzing populations of individual cells greatly simplifies the challenges of tissue cellular heterogeneity characterization and data correlation. By combining these two disparate chemically-rich modalities on the same single cell, we extend the information that can be gain to high resolution images with discrete lipid identities.

8.5 Acknowledgements

The authors gratefully acknowledge support from the National Institutes of Health,

Award Number 1U01 MH109062 from the National Institute of Mental Health. E.K.N. acknowledges support from the National Science Foundation Graduate Research Fellowship

Program and the Springborn Fellowship.

282

8.6 Figures and Table

Figure 8.1 Schematic of workflow combining single cell MALDI FT-ICR MS with SRSM.

Whole slide microscope images of populations of single cells are used to locate 9276 cell-like structures with microMS (A). Putative cells were determined object-to-object distances and object sizes filtering. 1342 targets were selected for MALDI FT-ICR MS analysis (B). Cells were then filtered by lipid abundance to reduce the sample size to 596 cells for SRSM analysis

(C). In total, 379 cells had high quality mass spectra and SRSM images.

283

Figure 8.2 Correlated optical, vibrational, and mass spectral single cell profiles from three animals. Morphological similarities between brightfield and SRSM indicate that the spectra belong to the same cell and that minimal morphological perturbation occurs after MALDI MS.

Single cell selected from animal 1, along with corresponding spectra (A). The mass spectrum

(A.i.) and vibrational spectrum (A.ii., rgb composite A.iii.) are shown for that specific cell. Cell selected from animal 2 (B) with resulting mass spectrum (B.i.) and vibrational spectrum (B.ii.,

284

Figure 8.2 (cont.) rgb composite B.iii.). Cell selected from animal 3 (C), with resulting mass spectrum (C.i.) and vibrational spectrum (C.ii., rgb composite C.iii.). This cell has lower lipid signal in comparison to the other two.

285

Figure 8.3 Mass spectrum of two cells that appear to be a single cell with brightfield microscopy, although are two distinct cells by SRSM.

286

Detected in Putative Assignment(s) Measured ppm Error Exact m/z Cell(s) m/z 1,2 [PC(32:4(OH)+H]+ 700.4547 700.454796 -0.1370 1,2 unknown 702.3019 1,2 unknown 703.3091 1,2 [PA(34:1+K]+ 713.4540 713.451814 3.0640 + 1,2,3 [PS(30:1)+NH4] 723.4920 723.491910 0.1244 1,2 unknown 725.2915 [SM(d36:1)+H]+ 1,2,3 + 731.6072 731.606151 1.4338 [SM(t36:0)+H-H2O] [PC(32:0)+H]+ 1,2,3 [PC(O-32:1(OH))+H]+ 734.5685 734.569432 -1.2688 [PC(P-32:0(OH))+H]+ [PA(36:2)+K 1,2 [PA(O-36:3(OH))+K]+ 739.4689 739.467464 1.9419 [PA(P-36:2(OH))+K]+ [PA(40:8)+H]+ 1,2,3 [PA(40:7(OH))+H- 745.4795 745.480282 -1.0490 + H2O] + 2,3 [MGDG(32:0)+NH4] 748.5901 748.593324 -4.3067 + 3 [PS(32:1)+NH4] 751.5239 751.523210 0.9181 [PA(38:2)+Na]+ 1,2,3 [PA(O-38:3(OH))+Na]+ 751.5253 751.524827 0.6294 [PA(P-38:2(OH))+Na]+ 1,2,3 [SM(d36:1)+Na]+ 753.5899 753.588096 2.3939 [PC(34:4)+H]+ 3 [PC(P-34:4(OH))+H]+ 754.5369 754.538131 -1.6315 [PC(O-34:3(OH))+H]+ 1,2,3 unknown 754.5407 1,3 unknown 754.5430 [PC(34:3)+H]+ 1,2,3 [PC(O-34:4(OH))+H]+ 756.5527 756.553782 -1.4302 [PC(P-34:3(OH))+H]+ [PC(34:1)+H]+ 1,2,3 [PC(O-34:2(OH))+H]+ 760.5855 760.585082 0.5496 [PC(P-34:1(OH))+H]+ + 1,2,3 [PE(40:7)+H-H2O] 772.5281 772.527567 0.6899 1,2 [PA(42:8)+H]+ 773.5149 773.511582 4.2895 [PE(38:1)+H]+ 3 [PE(O-38:2(OH))+H]+ 774.6028 774.600732 2.6698 [PE(P-38:1(OH))+H]+

Table 8.1 (cont.)

287

Detected in Putative Assignment(s) Measured ppm Error Exact m/z Cell(s) m/z [PC(36:4)+H]+ 1,2,3 [PC(O-36:5(OH))+H]+ 782.5683 782.569432 -1.4465 [PC(P-36:4(OH))+H]+ 1,2,3 [PC(36:1)+H]+ 788.6178 788.616382 1.7981 [PC(34:1)+K]+ 1,2,3 [PC(O-34:2(OH))+K]+ 798.5415 798.540963 0.6725 [PC(P-34:1(OH))+K]+ + [PI(O-34:2)+H-H2O] 1,2 + 803.5450 803.543276 2.1455 [PI(P-34:1)+H-H2O] 1,2,3 [PC(38:7)+H]+ 804.5562 804.553782 3.0054 [PC(38:5)+H]+ 1,2,3 [PC(O-38:6)+H]+ 808.5853 808.585082 0.2696 [PC(P-38:5)+H]+ [PC(38:4)+H]+ 1,2,3 [PC(O-38:5(OH))+H]+ 810.6002 810.600732 -0.6563 [PC(P-38:4(OH))+H]+ 2 unknown 819.5219 + 1,2 [PE(44:11)+H-H2O] 820.5281 820.527567 0.6496 2 [PI-Cer(t34:0)+Na]+ 820.5338 820.531034 3.3710 [PC(36:1)+K]+ 1,2 [PC(O-36:2(OH))+K]+ 826.5733 826.572263 1.2546 [PC(P-36:1(OH))+K]+ 1 [PI-Cer(t36:0)+H]+ 826.5796 826.580390 -0.9557 3 [PE(38:1(OH))+K]+ 828.5518 828.551528 0.3283 1,2,3 [PC(40:9)+H]+ 828.5557 828.553782 2.3149 [PI-Cer(d36:1)+Na]+ 1,2 830.5519 830.551770 0.1565 [PS(36:0(OH))+Na]+ [LacCer(d32:3)+H]+ 1,2 [LacCer(t32:2)+H- 830.562 830.562418 -0.5033 + H2O] 1,2,3 [PC(40:7)+H]+ 832.5860 832.585082 1.1026 1,2 unknown 838.3184 1,2 unknown 839.3275 3 unknown 834.5195 1,2,3 unknown 870.5883 1,2,3 unknown 896.6077 Table 8.1 (cont.) Assignments for lipid features detected in the three cells by MALDI FT-ICR

MS made using high mass accuracy (<5 ppm error) and the LIPID MAPS database.

288

8.7 References

(1) Shalek, A. K.; Satija, R.; Adiconis, X.; Gertner, R. S.; Gaublomme, J. T.; Raychowdhury,

R.; Schwartz, S.; Yosef, N.; Malboeuf, C.; Lu, D.; Trombetta, J. J.; Gennert, D.; Gnirke,

A.; Goren, A.; Hacohen, N.; Levin, J. Z.; Park, H.; Regev, A., Nature 2013, 498, 236.

(2) Tasic, B.; Menon, V.; Nguyen, T. N.; Kim, T. K.; Jarsky, T.; Yao, Z.; Levi, B.; Gray, L.

T.; Sorensen, S. A.; Dolbeare, T.; Bertagnolli, D.; Goldy, J.; Shapovalova, N.; Parry, S.;

Lee, C.; Smith, K.; Bernard, A.; Madisen, L.; Sunkin, S. M.; Hawrylycz, M.; Koch, C.;

Zeng, H., Nature Neuroscience 2016, 19, 335.

(3) Shekhar, K.; Lapan, S. W.; Whitney, I. E.; Tran, N. M.; Macosko, E. Z.; Kowalczyk, M.;

Adiconis, X.; Levin, J. Z.; Nemesh, J.; Goldman, M.; McCarroll, S. A.; Cepko, C. L.;

Regev, A.; Sanes, J. R., Cell 2016, 166 (5), 1308-1323.e30.

(4) Thomsen, E. R.; Mich, J. K.; Yao, Z.; Hodge, R. D.; Doyle, A. M.; Jang, S.; Shehata, S.

I.; Nelson, A. M.; Shapovalova, N. V.; Levi, B. P.; Ramanathan, S., Nature Methods

2015, 13, 87.

(5) Comi, T. J.; Do, T. D.; Rubakhin, S. S.; Sweedler, J. V., Journal of the American

Chemical Society 2017, 139 (11), 3920-3929.

(6) Ong, T.-H.; Kissick, D. J.; Jansson, E. T.; Comi, T. J.; Romanova, E. V.; Rubakhin, S. S.;

Sweedler, J. V., Analytical Chemistry 2015, 87 (14), 7036-7042.

(7) Comi, T. J.; Makurath, M. A.; Philip, M. C.; Rubakhin, S. S.; Sweedler, J. V., Analytical

Chemistry 2017, 89 (14), 7765-7772.

(8) Comi, T. J.; Neumann, E. K.; Do, T. D.; Sweedler, J. V., Journal of The American

Society for Mass Spectrometry 2017, 28 (9), 1919-1928.

289

(9) D., D. T.; F., E. J.; K., N. E.; J., C. T.; G., T. E.; E., L. A.; S., R. S.; V., S. J.,

ChemPhysChem 2018, 19 (10), 1180-1191.

(10) Kompauer, M.; Heiles, S.; Spengler, B., Nature Methods 2016, 14, 90.

(11) Zavalin, A.; Yang, J.; Hayden, K.; Vestal, M.; Caprioli, R. M., Analytical and

Bioanalytical Chemistry 2015, 407 (8), 2337-2342.

(12) Rubakhin, S. S.; Garden, R. W.; Fuller, R. R.; Sweedler, J. V., Nature Biotechnology

2000, 18, 172.

(13) Rubakhin, S. S.; Greenough, W. T.; Sweedler, J. V., Analytical Chemistry 2003, 75 (20),

5374-5380.

(14) Wei, L.; Hu, F.; Shen, Y.; Chen, Z.; Yu, Y.; Lin, C.-C.; Wang, M. C.; Min, W., Nature

Methods 2014, 11, 410.

(15) Fu, D.; Zhou, J.; Zhu, W. S.; Manley, P. W.; Wang, Y. K.; Hood, T.; Wylie, A.; Xie, X.

S., Nature Chemistry 2014, 6, 614.

(16) Cheng, J.-X.; Xie, X. S., Science 2015, 350 (6264).

(17) Lanni, E. J.; Dunham, S. J. B.; Nemes, P.; Rubakhin, S. S.; Sweedler, J. V., Journal of

the American Society for Mass Spectrometry 2014, 25 (11), 1897-1907.

(18) Gelber, M. K.; Kole, M. R.; Kim, N.; Aluru, N. R.; Bhargava, R., Analytical Chemistry

2017, 89 (3), 1716-1723.

(19) Freudiger, C. W.; Min, W.; Saar, B. G.; Lu, S.; Holtom, G. R.; He, C.; Tsai, J. C.; Kang,

J. X.; Xie, X. S., Science 2008, 322 (5909), 1857-1861.

(20) Neumann, E. K.; Comi, T. J.; Spegazzini, N.; Mitchell, J. W.; Rubakhin, S. S.; Gillette,

M. U.; Bhargava, R.; Sweedler, J. V., Analytical Chemistry 2018.

(21) Piomelli, D., Prostaglandins & Other Lipid Mediators 2005, 77 (1), 23-34.

290

(22) Yang, K.; Han, X., Trends in Biochemical Sciences 2016, 41 (11), 954-969.

(23) Piomelli, D.; Astarita, G.; Rapaka, R., Nature Reviews Neuroscience 2007, 8, 743.

(24) Wenk, M. R., Nature Reviews Drug Discovery 2005, 4, 594.

(25) Wu, H.; Volponi, J. V.; Oliver, A. E.; Parikh, A. N.; Simmons, B. A.; Singh, S.,

Proceedings of the National Academy of Sciences 2011, 108 (9), 3809-3814.

(26) Passarelli, M. K.; Ewing, A. G., Current Opinion in Chemical Biology 2013, 17 (5), 854-

859.

(27) Page, J. S.; Rubakhin, S. S.; Sweedler, J. V., Analytical chemistry 2002, 74 (3), 497-503.

291

CHAPTER 9

INTERROGATION OF SPATIAL METABOLOME OF GINKGO BILOBA WITH

HIGH‐RESOLUTION MATRIX‐ASSISTED LASER DESORPTION/IONIZATION AND LASER

DESORPTION/IONIZATION MASS SPECTROMETRY IMAGING

Notes and Acknowledgements

This chapter is adapted from a completed manuscript coauthored by Bin Li, Junyue Ge,

Wen Gao, Hua Yang, Ping Li, and Jonathan V. Sweedler, submitted for publication on February

6, 2018 and accepted June 27, 2018 in Plant, Cell & Envin Prof. B. Li wrote a majority of the manuscript, performed sample preparation, liquid chromatography-MS analysis, MALDI-time- of-flight-mass spectrometry, most data analysis, and conceived the concept. J. Ge, W. Gao, and

H. Yang aided in sample preparation. Prof. P. Li and J.V. Sweedler funded the work. B. Li gratefully acknowledges financial support from the National Natural Science Foundation of

China (no. 81773873). P. Li acknowledges support from the 111 Project (no. B16046).

J.V. Sweedler acknowledges support from the Center for Advanced Bioenergy and Bioproducts

Innovation funded by the U.S. Department of Energy. E.K. Neumann acknowledges support from the U.S. National Science Foundation Graduate Research Fellowship Program and the

Springborn Fellowship. We also thank Weiwei Tang for his assistance in creating histogram graph. The authors gratefully acknowledge support from the National Institutes of Health, Award

Number P30 DA018310 from the National Institute on Drug Abuse, and from the National

Institute of Mental Health, Award Number 1U01 MH109062.

292

9.1 Introduction

Ginkgo biloba is an ancient plant, the only extant species in the division Ginkgophyta, and therefore is considered to be a “living fossil”.[1] Because of its high economic value in the food, pharmaceutical, and cosmetic industries, G. biloba continues to receive much attention in scientific investigations.[2] In phytochemistry, a wide variety of bioactive metabolites have been isolated and identified from different G. biloba tissues, for example, roots, stems, leaves, and seeds.[1] Among them, ginkgo leaves are a productive chemical factory where diverse secondary metabolites are produced that have significant pharmacological activities.[3] According to clinical research, ginkgo leaf extracts have the potential for treatment of diseases associated with peripheral circulation, memory, and cognitive dysfunction.[4] Additionally, G. biloba is remarkable because of its high environmental adaptability and unparalleled tolerance of environmental stress.[5] It is a unique species representative of a whole branch of the phylogenetic tree of seed plants.

Investigations employing modern plant metabolomic techniques have led to the comprehensive profiling of metabolites from the ginkgo plant. For example, gas chromatography

(GC), liquid chromatography (LC), capillary electrophoresis (CE), hyphenated methods such as

GC/LC/CE–mass spectrometry (MS), and nuclear magnetic resonance have enabled the analysis of large numbers of metabolites.[3] However, the spatial localization of metabolites in samples is often lacking because the analyses are performed on tissue homogenates. Generally, plants comprise at least 10 basic tissue types, with more than 15 structurally diverse cell types.[6] The metabolism in plant tissues is often asymmetrically distributed among different cellular and subcellular compartments, resulting in the heterogeneous distribution of plant metabolites.

Therefore, it can be important to unravel and understand the spatial organization of metabolites

293 in different tissues to gain fundamental insights into the physiological, growth, development, and stress responses of plants. Currently, plant imaging techniques, such as light and electron microscopy, are used for morphological characterization and chemical localization of labelled components in plant tissues. However, there are significant limitations using these techniques as they require labelling of specific chemicals, limiting de novo investigation of unlabeled and unknown chemical compounds.

MS imaging (MSI) has become an attractive molecular imaging tool for capturing the spatial distribution of plant metabolites.[7-11] It is label‐free, untargeted, multiplexed, and capable of de novo chemical discovery. Various MSI approaches have been applied to image metabolites within tissues. For example, matrix‐assisted laser desorption/ionization (MALDI) imaging,[12, 13] secondary ion mass spectrometry (SIMS),[13] and ambient ionization‐based MSI techniques such as desorption electrospray ionization (DESI)[14] and laser ablation electrospray ionization

(LAESI).[15, 16] It is important to select the method most appropriate for the system of interest, because each approach has differences in analyte coverage, sample preparation, and spatial resolution. For example, SIMS can achieve high lateral resolution, generally between 50 and

1,000 nm, to examine molecules and molecular fragments below an m/z value of 1,000. DESI requires minimal sample preparation but has a spatial resolution limited to around 100 μm.

MALDI is currently the most extensively used MSI ionization method because of its broad analyte coverage and satisfactory spatial resolution, ranging from 10 to 100 μm using commercial instruments; however, a custom‐built laboratory instrument capable of 1.4 μm resolution has recently been reported.[17] In MALDI MSI, the type of matrix and the method of matrix application need to be optimized because these steps determine analyte coverage, limits of

294 detection, ionization efficiency, spatial resolution, and the ultimate quality of the ion images.[10,

18]

In this work, spatio‐chemical information on metabolites in a cross section of ginkgo leaf and in different organs has been obtained using high‐resolution Fourier transform ion cyclotron resonance (FT‐ICR) MALDI and laser desorption/ionization (LDI) MSI. In particular, sample preparation directly associated with the quality and authenticity of the MSI results was optimized for the ginkgo leaf. Numerous constituents, including flavonoid aglycones, biflavonoids, flavonoid glycosides, biginkgosides, ginkgolides, and phenolic lipids, were comprehensively visualized in ginkgo leaf, stem, and root for the first time. High mass accuracy and in situ tandem

MS (MS/MS) measurements helped in identifying metabolites within the tissue sections. The spatial information obtained here suggests functional roles for metabolites, as well as hints at localized responses in the leaf to biotic and abiotic stressors.

9.2 Materials and Methods

Chemicals and plant samples

Formic acid (LC–MS grade), water (LC–MS grade), acetonitrile (ACN; LC–MS grade), trifluoroacetic acid (TFA; LC–MS grade), 2,5‐dihydroxybenzoic acid (DHB),

α‐cyano‐4‐hydroxycinnamic acid (CHCA), and 9‐aminoacridine (9‐AA) were purchased from

Sigma‐Aldrich (St. Louis, MO, USA). G. biloba leaves were collected from the arboretum of

University of Illinois at Urbana‐Champaign, USA. Additionally, 2‐year‐old G. biloba plant grown in botanical garden at China Pharmaceutical University, Nanjing, China, was selected for

MSI measurement of ginkgolides in root, young stem, and leaf tissue.

295

Sample preparation for MALDI imaging

Fresh G. biloba leaves, young stems, and root parts were immediately embedded in 10% gelatin (wt/vol) solutions. Initially, tissues were kept in Tissue‐Tek cryomolds (25 × 20 × 5 mm) and the gelatin solution was poured over them to embed the tissues; thereafter, the molds were transferred to a −80°C freezer for 30 min to form a solid block. For cryosectioning, the sample blocks were directly fixed on the sample holder of a cryostat (Leica, Germany), using deionized water as the adhesive. Sections of 16 μm thickness were obtained at −20 °C and thaw‐mounted on indium tin oxide‐coated glass slides for immediate imaging measurements. To avoid condensation, the tissue sections were dehydrated in a vacuum desiccator for approximate

10 min prior to matrix application. A Zeiss Axio M2 microscope (Zeiss, Germany) was used to obtain optical images of the sections.

A laboratory‐constructed automated pneumatic‐assisted matrix application system was used for the uniform application of MALDI matrix solution. The matrix application system and coating procedure were similar to previously published work with some modifications.[19]

Briefly, 50 mg ml−1 DHB or 15 mg ml−1 CHCA dissolved in ACN:water (0.1% TFA; 7:3, vol/vol) was applied for the positive mode MALDI experiments. For negative mode MALDI,

10 mg ml−1 9‐AA dissolved in methanol:water (9:1, vol/vol) was applied. For homogenous deposition onto the leaf samples, the nebulizer was held 3 cm above the sample and oscillated over the plate 100 times. The flow rate was set to 6–8 ml hr−1 and gas pressure to 50 psi, to deliver and nebulize the matrix solution, respectively. Additionally, DHB was sublimed using a home‐built apparatus described in detail elsewhere.[20] Briefly, 1 g of powdered DHB was distributed evenly on an aluminium foil boat, and the samples were affixed to a copper adaptor on the outside of the cold finger. The chamber was closed, placed in a heating mantle (Model

296

No. 0408, Glas‐Col, Terre Haute, IN, USA), and pumped to intermediate vacuum (~10 mTorr), and the cold finger was filled with iced water. After 2 min of temperature equilibration, the desired matrix coating was achieved by supplying 120 V to the heating mantle for 9.5 min for

DHB at 55% power with a variable autotransformer (Staco Inc., Dayton, OH, USA). The chamber was removed from the mantle and vented to room temperature air, and the sample was promptly removed from the cold finger.

MALDI MSI instrumentation and analysis parameters

MALDI imaging was performed using a 7T solariX FT‐ICR mass spectrometer (Bruker

Corp., Billerica, MA) equipped with a dual ESI/MALDI source and a Smartbeam II laser.

An m/z range of 150–2,000 was acquired at 8 Mword, producing a transient length of 5.8720 ms.

It should be noted that the length of the transient can be shortened depending on the specific analytes measured. Single‐scan spectra consisted of 100 accumulated laser shots at 1 kHz with laser focus set to “small.” MALDI images were acquired at a 50‐μm spatial resolution.

Instrument calibration was performed using the ESI portion of the instrument with NaTFA clusters ranging from m/z 158.9640–1,926.6365 and a quadratic calibration of the clusters.

Because the ESI and MALDI generated ions are transferred to and are analyzed with the same ion optics and detection systems, the calibration remains valid for ions generated with both ion sources. Data were analyzed using Data Analysis version 4.0 and flexImaging version 4.1 software (Bruker). Additionally, initial experiments to determine the choice of matrix and some of the in situ MS/MS analyses were conducted using an ultrafleXtreme MALDI time‐of‐flight

(ToF)/ToF mass spectrometer with a frequency tripled Nd:YAG solid‐state laser (Bruker Corp.).

297

Sample preparation for LC–MS/MS analysis

The upper and lower epidermal cell layers were manually dissected from leaf cross sections (30 μm thickness) using sharp tungsten needles and transferred into Eppendorf tubes.

All of the epidermis isolation steps were carried out under a stereomicroscope (Leica). Each sample was collected from five leaf sections and combined for subsequent solvent extraction;

100 μl of methanol:water (1:1, vol/vol) was added into the tubes, followed by 15 min of sonication (8891 ultrasonic cleaner, Billerica, MA). The extracts were centrifuged at 9600 × g rpm for 5 min, and the supernatants were collected. The extract solutions were then dried with a gentle N2 stream at ambient conditions and room temperature. The dried residuals were reconstituted in 20 μl methanol:water (1:1, vol/vol) for subsequent LC–MS/MS analysis.

Relative quantitation of selected flavonoids by LC–MS/MS

LC–MS/MS was performed using an EVOQ Elite™ Triple Quadrupole mass spectrometer (Bruker Corp.) connected to an Advance UHPLC system (Bruker Corp.). Analytes were separated on a Fortis UHPLC C18 column (1.7 μm, 50 mm × 2.1 mm, Fortis Technologies

Ltd., UK). Mobile phase A was H2O containing 0.1% formic acid and phase B was ACN. The gradient programme was conducted as follows: 0–15 min, 5–95% B; 15–16 min, 95% B; 16–

16.1 min, 95–5% B; and 16.1–18 min, 5% B. Total run time was 18 min. The injection volume was 2 μl. The mass spectrometer was operated in negative ion mode, and the EVOQ source parameters were as follows: heated ESI, spray voltage (−) 4,500 V; cone temperature, 250°C; cone gas flow, 25; heated probe temperature, 450°C; probe gas flow, 45; nebulizer gas flow, 65; and exhaust gas, off. Argon was used as a collision gas in the multiple reaction monitoring

(MRM) mode. The MRM parameters, including ion transitions (precursor and product ions) for

298 each analyte, are presented in Table 9.1. EVOQ MRM chromatograms were analyzed using Data

Review 8.2 (Bruker).

The content of each selected analyte was measured from three replicate leaf samples. For comparison of averaged peak intensities of each target, analytes were acquired from the upper and lower epidermis, respectively; a one‐tailed Student's t‐test was calculated by Originpro 8.5

(Northampton, MA, USA) to determine the mean differences in flavonoid content between the upper and lower epidermis.

Identification of Biginkgosides

Our initial database search, METLIN and KNApSAcK with accurate mass match did not reveal reliable chemical identities. With the aid of MS/MS, important hints for the deduction of possible chemical structures were obtained. They were tentatively attributed to the dimer of flavonoid glycosides. For example, for m/z 1535.3476, the ion formula was calculated as

+ [C72H72O35 + K] with –0.66 ppm mass accuracy, and the characteristic fragment ions were obtained by MALDI CID-FT-ICR MS/MS. The fragment ion at m/z 1249.30 was assigned to the loss of the kaempferol group (286.05 Da), m/z 1233.30 assigned to the loss of the quercetin group (302.05 Da), and m/z 947.26 assigned to the loss of the kaempferol and quercetin groups

(588.09 Da). Additionally, two minor fragment peaks at m/z 795.14 and m/z 779.15 were tentatively assigned to the loss of kaempferol-coumaroylglucosyl-rhamnoside (740.20 Da) and quercetin-coumaroylglucosyl-rhamnoside (756.19 Da). Therefore, m/z 1535.3476 was tentatively assigned to the potassium adduct of a dimer of kaempferol-coumaroylglucosyl-rhamnoside and quercetin-coumaroylglucosyl-rhamnoside. Likewise, m/z 1519.3557 and m/z 1551.3435 could be tentatively assigned to the potassium adduct of a dimer of kaempferol-coumaroylglucosyl-

299 rhamnoside and the potassium adduct of a dimer of quercetin-coumaroylglucosyl-rhamnoside, respectively (Figure 10.2). However, it is hard to determine the origin of these dimers, which are possibly derived from native endogenous metabolites or occur as a result of the measurement process; for example, lipid dimers are often observed after laser irradiation in MALDI MS.

Identification of Lipids

Peak assignment of lipid-associated ions was carried out by accurate mass match and/or characteristic fragments aided by Lipid Maps and the METLIN database. As shown in Figure

9.1D, numerous dominant ions with an even mass number were putatively assigned to the phosphatidylcholine (PC) lipids according to the predominant diagnosed fragments resulting from the loss of the polar choline phosphate head group (183 Da) and/or quaternary ammonia group (59 Da, N(CH3)3). For the other lipid-associated ions, in situ MS/MS was not available and so they were tentatively assigned according to accurate mass match (Table 9.2).

Identification of Chlorophyll a

In Figure 9.1D, a distinguished peak at m/z 871.5724 observed in mesophyll cells was ascribed to [chlorophyll a-Mg+3H]+ according to the accurate mass match and previously reported data. The loss of the central magnesium ion was due to the use of acidic matrices, e.g.

DHB, which are known to facilitate the preferential demetalation of magnesium porphyrins. The identity of the peak at m/z 871.5724 was further demonstrated by other characteristic ions such as m/z 870.5700, corresponding to [chlorophyll a-Mg+2H]+•, and m/z 593.2759, originating from the loss of a phytyl residue from [chlorophyll a-Mg+ 3H]+.

UV-B radiation Plant materials

300

12 leaves were taken from 2-year old ginkgo plants grown under same environmental conditions in the botanical garden at China Pharmaceutical University, Nanjing, China. 6 leaves were chosen as the control group, and 6 leaves were selected to conduct UV-B radiation experiments. UV-B treatment was carried out in UV-B chambers (Nanjing Kenfan Electronic

Technology Co Ltd, China) with UV-B lamps emitting approximately 36 W of UV-B irradiation with 313 nm. The UV-B intensity was 225 μW·cm-2 determined by a UV radiometer

(Photoelectric Instrument Factory of Beijing Normal University, China).

In UV-B radiation experiment, the whole plant was placed in the chamber irradiated for 4 hours. After an adaptation time of 24 h, leaf samples were picked for subsequent LC-MS/MS analysis. For control group, leaf samples were directly collected before UV-B radiation for subsequent LC-MS/MS analysis. Each leaf sample was extracted with 1 mL of methanol, followed by 15 min of sonication. The extracts were centrifuged at 13,000 rpm for 5 min and the supernatants were collected. The 90 µL supernatant added with the 10 µL internal standard solution (naringin, 0.27 mg/mL) was filtrated through a 0.22 mm membrane prior to LC MS/MS analysis.

Relative quantitation of selected flavonoids by LC-MS/MS

LC-MS/MS analysis of samples with and without UV-B exposure was performed using an Agilent 6460 Triple Quadrupole mass spectrometer connected to Agilent series1290 Infinity

HPLC system (Agilent Technologies, USA). Analytes were separated on an Agilent ZORBAX

Extend C18 column (4.6 × 50 mm, 1.8 μm). Mobile phase A was H2O containing 0.1% formic acid and phase B was acetonitrile. The gradient program was conducted as follows: 0–15 min, 5–

95% B; 15–17 min, 95% B; 17–17.1 min, 95–5% B; 17.1–18 min, 5% B. Total run time was 18 min. The injection volume was 2 μL. The mass spectrometer was operated in the negative ion

301 mode and the source parameters were as follows: drying gas temperature, 300 °C; drying gas flow, 5 L/min; nebulizer pressure, 45 psi; capillary voltage, 3500 V. N2 was used as a collision gas in the multiple reaction monitoring (MRM) mode. The MRM parameters, including ion transitions (precursor and product ions) for each analyte, are presented in Table 9.1. Data acquisition was performed with Mass Hunter Workstation (Agilent Technologies, USA).

The content of each selected analyte was measured from 6 replicate leaf samples from the control group and UV-B treated group. For comparison of averaged peak intensities of each target, analytes were acquired from two groups, respectively; a one-tailed Student’s t test was calculated by Originpro 8.5 (Northampton, MA, USA) to determine the mean differences in flavonoid content between the control group and UV-B treated group.

9.3 Results

Characterizing metabolite profiles of ginkgo leaf using MALDI and LDI MS

Initially, leaf cross sections were probed by MALDI‐ToF MS to determine the optimal

MALDI matrix. Several matrices (DHB and CHCA for positive ion mode and 9‐AA for negative ion mode), the amount of matrix applied onto the sample, and two matrix application approaches

(automatic sprayer and sublimation) were assessed for achieving comprehensive detection of metabolites within the ginkgo leaf. Briefly, biflavonoids, flavonoid glycosides, and lipids were more readily detected with DHB compared with CHCA in positive mode, with few flavonoids detected with 9‐AA in negative mode (Figure 9.2). Therefore, DHB was selected for analyzing ginkgo leaf sections in positive ion mode.

Low molecular weight compounds, such as flavonoid aglycones, were not readily resolved using MALDI‐ToF MS due to interfering DHB matrix peaks in the low mass range

302

(

Additionally, LDI was used to detect the ultraviolet (UV)‐absorbing flavonoids within the ginkgo leaf. As expected, ions associated with flavonoid aglycones (Figure 9.3A) and biflavonoids (Figure 9.3B) were observed in negative mode LDI MS; however, few flavonoid glycosides were detected. Interestingly, many unexpected, tissue‐specific ions were exclusively detected in the secretory cavities of ginkgo leaves with negative mode LDI MS (Figure 9.3C), which were subsequently identified as ginkgolic acids and cardanols according to their accurate masses and MS/MS analysis (Figures 9.1F).

Representative single‐pixel matrix‐assisted laser desorption/ionization (MALDI) and laser desorption/ionization (LDI) Fourier‐transform ion cyclotron resonance mass spectra acquired from a cross section of ginkgo leaf. (A, B, C, and E) Positive (POS) mode MALDI mass spectrum of flavonoids obtained from upper epidermis. (D) Positive mode MALDI mass spectrum of lipids and chlorophylls obtained from mesophyll. (F) Negative (NEG) mode LDI mass spectrum of ginkgolic acids and cardanols obtained from secretory cavity. (G) Negative mode LDI mass spectrum of flavonoid aglycones obtained from upper epidermis. Identified compounds are labelled with measured mass. See Tables 9.3-6 for more details.

Additionally, two matrix application approaches, air‐assisted sprayer (wet coating) and sublimation (dry coating), were compared for optimal signal intensity and analyte coverage, as well as minimization of metabolite delocalization. As shown in Figure 9.4, biflavonoid‐associated ions were detected from the epidermis with both application approaches; the flavonoid glycosides were rarely observed after sublimation. Therefore, a laboratory‐constructed air‐assisted sprayer was used to provide homogenous coating across the leaf cross sections.[19]

303

Although high accuracy mass peaks are often detected in MSI, in situ assignment of these peaks to specific metabolites remains a challenge for MSI. Here, we facilitated assignments by performing follow‐up measurements with high‐resolution FT‐ICR MS and in situ MS/MS, accomplished via either ToF/ToF MS/MS or collision‐induced dissociation MS/MS using the

FT‐ICR mass spectrometer (Tables 9.3–9.6). Figure 9.1 shows representative MALDI and LDI

FT‐ICR mass spectra obtained from a cross section of ginkgo leaf. Most ions corresponding to flavonoid aglycones were sufficiently resolved from matrix ions (Figure 9.1A). In positive ion mode, flavonoid aglycones and biflavonoids were readily detected as protonated ions; however, flavonoid glycosides were mainly detected as sodium‐ and/or potassium‐adduct ions (Table 9.3).

More detailed structural information was subsequently obtained using MS/MS. However, in situ

MS/MS was not always possible because some ions were too low in abundance, or characteristic fragment ions were not detected. Furthermore, as shown in Figure 9.1G, three pairs of characteristic flavonoid aglycones peaks were observed, for example, m/z 284.0331 and

285.0409, m/z 300.0279 and 301.0359, m/z 314.0437 and 315.0517. These are deprotonated ions

[M−H]− and radical product ions [M−H−H]−• due to the presence of a free 4′‐OH group on the

B‐ring of some aglycones.[21]

Interestingly, as shown in Figure 1e, three intense peaks around m/z 1,500 were exclusively observed in the epidermis. They were tentatively attributed to the dimer of flavonoid glycosides, after high accurate mass and MS/MS information was considered. For

+ example, m/z 1,519.3557 (calculated for [C72H72O34 + K] , Δ = 1.34 ppm) was putatively assigned as a dimeric derivative of kaempferol coumaroyl diglycoside (C36H36O17). In a recent report, a new class of flavonoid glycoside cyclodimers, named biginkgosides, were isolated from ginkgo leaf extract, supporting our MALDI MS results. Therefore, m/z 1,519.3557 was

304 tentatively assigned to isomeric biginkgosides ([biginkgoside B/E/I + K]+), m/z 1,535.3476 assigned to [biginkgoside F + K]+, and m/z 1,551.3435 assigned to the other isomeric biginkgosides also adducted with potassium ([biginkgoside A/C/D/G/H + K]+; Table 9.4).

In Figure 9.1F, a series of ions was observed in the low mass range around m/z 300 with a specific mass shift (e.g., m/z 26.0156 (CH═CH) and m/z 28.0313 (CH2─CH2)), suggesting the possible occurrence of lipid‐associated ions. They were only observed in the secretory cavities from the leaf section. Combined high accurate mass and characteristic fragment ions obtained with LIFT™ ToF/ToF MS/MS, ions with a common loss of 44.0 Da (CO2) were assigned to

−• ginkgolic acids (GAs), whereas ions with a common fragment of m/z 105.8 ([C7H6O] ) were assigned to cardanols.[22] In total, seven GAs with different alkyl side chains and degrees of unsaturation were identified, including GA (C13:0, C15:2, C15:1, C15:0, C17:3, C17:2, and

C17:1) (Table 9.5). Four cardanols were also identified, including cardanol (C13:0, C15:1,

C15:0, and C17:1) (Table 9.5). Besides the above‐mentioned metabolites, ginkgolides, saccharides, various phospholipids, and chlorophyll a (m/z 871.5724, [chlorophyll a‐Mg + 3H]+) were also detected in leaf cross sections. Peak assignment of these ions is summarized in the

Compound Identification section of the experimental and Table 9.6).

MALDI/LDI MSI reveals heterogeneous distribution of metabolites in the ginkgo plant

Ginkgo leaves are fragile tissues of less than 1 mm thick, making direct cryosectioning challenging, especially while maintaining tissue integrity and avoiding metabolite delocalization.

In this work, fresh leaf tissues were embedded in gelatin solutions before cryosectioning to improve the cross sections available for MSI (Figure 9.5). The optical image in Figure 9.5 of a leaf cross section shows the main compartments: epidermis, vascular bundle, mesophyll, and

305 secretory cavity. Summaries of the distinctive spatial distributions of various kinds of metabolites are shown in Figures 9.6, 9.7, and 9.8. Figure 9.96 shows two‐dimensional ion intensity maps of flavonoid aglycones, biflavonoids, flavonoid glycosides, and biginkgosides, which mainly accumulate in the upper and lower epidermis. GAs and cardanols were exclusively detected within highly specialized cell compartments—secretory cavities (Figure 9.7). The distributions of ginkgolide A (m/z 447.1052) and a group of isomers of ginkgolide

(m/z 463.1001; ginkgolide B/J/M/K) are located within both the epidermis and mesophylls as compared with flavonoids (Figure 9.8), which were only detected in the epidermis. High abundances of saccharides and lipid‐associated ions were observed mostly in the mesophyll layers (Figure 9.8). Several m/z values specifically found in the secretory cavities were tentatively assigned to lipid‐associated ions according to accurate mass match in the METLIN database; however, further work is required to identify these ions (Figure 9.9). As expected, chlorophyll a was visualized within chloroplasts, which are responsible for the characteristic green color in plants. Additionally, organ‐specific distributions of ginkgolides belonging to the terpene lactones were observed. As shown in Figure 9.10, ginkgolide A, B, and/or other isomers are mainly detected in the cortex and phloem region of root and mesophylls of leaf. The highest amount of ginkgolides is detected in the root tissue and decreased in the order of the leaf and young stem.

LC–MS validation of uneven distribution of major flavonoids in epidermis

As shown in Figure 9.4, a non‐uniform distribution of flavonoids exists between the upper and lower leaf epidermis, indicating a higher abundance of flavonoids in the upper leaf epidermis. To validate this difference in distribution observed in MALDI MSI, the upper and lower epidermal cell layers were dissected separately from leaf cross sections (free of mesophyll,

306 inset in Figure 9.11); thereafter, extraction and LC–MS/MS analysis were performed for relative quantitation of the content of each selected flavonoid (Table 9.1). As shown in Table 9.1, nine compounds belonging to flavonoid aglycone, biflavonoids, and flavonoid glycosides, respectively, were selected to validate the imaging results. Figure 9.11 shows the mean content of selected flavonoids between the upper and lower epidermis. Statistical analyses of the mean difference showed that the relative amount of each selected flavonoids was significantly higher at the upper epidermal layer compared with the lower side (≥99.5 confidence level), which correlated well with the distribution pattern observed in the MALDI imaging results.

9.4 Discussion

Although many experiments have been focused on the characterization and functional analysis of bioactive components in G. biloba, their spatial distributions oftentimes remain unresolved. Unlike well‐established approaches for visualization of morphological structures and localization of macromolecules such as genes and enzymes, precisely determining the spatial distribution of individual metabolites using optical techniques is difficult when most of the compounds are uncharacterized or unknown. Therefore, unambiguous mapping of the specific sites of production, transportation, and accumulation of many metabolites in tissues has yet to be realized. Here, we demonstrate that MALDI and LDI FT‐ICR MSI are well suited for investigating spatial metabolome of ginkgo leaf. Additionally, to validate the observed differences in metabolites between the upper and lower epidermis, and to verify that the observed distributions are not sample preparation artefacts, we performed LC–MS/MS

(Figure 7). The LC–MS/MS data demonstrated a similar significant difference in the abundance of detected flavonoids between the upper and lower epidermis.

307

One challenge we encountered was preparing high‐quality cryosections from the leaf for subsequent MSI measurements. As shown in Figure 9.4, ginkgo leaf tissues have complex and intricate structures, such as stomas, a loosely packed mesophyll layer, and large intercellular spaces and resin canals, which must be intact after sectioning for accurate metabolite mapping.

Tissue embedding is an efficient way to preserve plant tissues in their original form and structure; however, the most common type of embedding using optimal temperature cutting material is incompatible with MSI as it strongly suppresses other ions.[23, 24] As such, we tested a variety of conditions (data not shown) and found that a 10% gelatin solution allowed for sectioning with minimal distortion or deterioration of the tissue. With gelatin embedding, we were able to section tissues between 12 and 16 μm, which provided an effective tissue thickness, overall ion intensity, and laser focus for MSI.

Figure 9.6 displays the distribution patterns of four different classes of flavonoids: aglycones and related glycosides, biflavonoids, and biginkgosides. The distribution of the flavonoids shared some similarities; for example, they were colocalized within the epidermis with higher abundance on the upper side. As previous shown, the heterogeneous distribution of chalcone synthase bioactivity (a key enzyme in flavonoid biosynthesis) was investigated in the leaf of the Pisum mutant; results showed that the majority of chalcone synthase activity was present exclusively in the epidermis, and the total activity measured in the upper epidermis was quite high compared with the lower side.[25] Flavonoids have been reported to protect plants from

UV‐B radiation[26] and to act as a chemical barrier to herbivores.[27] Here, a UV‐B radiation experiments were performed to investigate the change of flavonoids in ginkgo leaves quantified using LC–MS/MS. As shown in Figure 9.12, after 4‐hr radiation, an increasing trend of target

308 flavonoids was observed, indicating the ginkgo leaves can protect themselves from radiation exposure by increasing UV‐absorbing flavonoids.

We also detected flavonoid cyclodimers that were localized to the epidermal region of the gingko leaf (Figure 9.6). Biginkgosides B, E, and/or I (m/z 1,519.3557) and biginkgoside F (m/z

1,535.3476) were mainly located in the upper epidermis, whereas biginkgoside A, C, D, G, and

H (m/z 1,551.3435) was detected in both the upper and lower epidermis, despite being from the same chemical class. Classical staining approaches (e.g., AlCl3 solution) would be unable to detect this difference, further demonstrating the utility of MSI. Although the roles of these flavonoid cyclodimers have yet to be determined, the distribution patterns obtained here for the first time may be significant for understanding what biotic factors trigger the production of these unusual dimeric derivatives, their corresponding heterogeneous accumulation patterns, as well as their function in ginkgo leaf development. Investigations of natural products in ginkgo leaf have been performed using effective dereplication strategies for a long time, but these flavonoids were only just recently reported.[28] One possible explanation of their recent discovery and observation here is that rare components with low abundance may not be observed during whole sample homogenization but amendable to MSI because of their localized nature. MALDI MSI is well suited for dereplication of spatially localized novel metabolites from plants, especially considering its minimal sample preparation steps.[29, 30]

Besides being a flavonoid‐rich resource, gingko leaf also has toxic phenolic lipids (e.g.,

GAs and cardanols), which can cause allergic reactions.[2] Therefore, the United States

Pharmacopeia prescribes an upper limit for the GA content of G. biloba extract. Using negative mode LDI MS, we detected these toxic phenolic lipids exclusively in the secretory cavities of ginkgo leaf (Figure 9.4). The mature cavity appears round and as a lipophilic droplet embedded

309 in mesophyll tissues. In situ MS/MS analysis of these compounds was not possible because of their low abundance and small sampling area; therefore, a sharp glass capillary was used to extract these lipophilic droplets directly onto a MALDI 384 target plate for LDI MS/MS analysis. With this complementary method, the identities of phenolic lipids in the secretory cavities were verified, validating the MALDI imaging results. Moreover, signals related to phenolic lipids were not observed in other parts of the leaf section, indicating that there was no redistribution of lipophilic droplets because of the steps taken to ensure optimal sample preparation.

Phenolic lipids, such as anacardic acids and cardanols, have been shown to inhibit bacterial, fungal, protozoan, and parasite growth, indicating that they may play important roles in chemical defense mechanisms.[31] Interestingly, previous studies have demonstrated that resin ducts in the secondary phloem of Rhus glabra L.,[32] the shoot of Mangifera indica L.,[33] and the root of Thapsia garganica[34] also store lipophilic components with the toxic properties. In our study, toxic phenol lipids were localized in the secretory cavities of ginkgo leaf tissue. This commonality found between the different plant families suggests that the secretory structures function as a pool to store the toxic metabolites, which can be released upon leaf crush to protect against herbivory.[35]

Another significant metabolite class is the terpene lactones (e.g., ginkgolides), here mainly localized to cortex and phloem region of root tissue and leaf epidermis and mesophyll

(Figure 9.7 and 9.8). Of course, accurate mass measurements do not differentiate isomers, such

+ as ginkgolide B, J, M, and K (C20H24O10, [M + K] , m/z 463.1001). Although in situ MS/MS is an effective approach for isomeric identification, characteristic fragments are often not detected because of the low abundance of the parent ion. High‐resolution ion mobility MS would be an

310 alternative option for separating isomeric metabolite ions based on their mobility in a high electrical field.[36] The spatial distribution of ginkgolides visualized with MSI is consistent with previous research performed using various techniques, such as isotope labelling, gene expression analysis, and tissue‐specific metabolite profiling.[37, 38] These prior results suggest that the site of ginkgolide biosynthesis is localized in the root, and the products are then translocated and accumulated mainly in the leaf tissues. In situ interrogation of spatial distribution of metabolites with MSI provides more straightforward and rich information; the MSI results are in agreement with the above‐mentioned methods. However, because of the complex interplay between biosynthesis and transport in the plant, it was not immediately clear in which organ or tissue metabolites are synthesized using only MSI. Therefore, a combination of multimodal imaging such as MSI and in situ hybridization is an effective method for the precise interpretation of complex biosynthesis of natural products.[39] Additionally, as the ginkgolides are a class of pharmacologically active ingredients with significant clinical applications, the imaging results obtained here may provide insights on how to increase the accumulation of bioactive ginkgolides in leaf tissue via abiotic stress or establish a high‐yield cell culture system.

Owing to the inherent capability of MSI for untargeted imaging, the spatial distribution of carbohydrates, phospholipids, and chlorophyll a were also revealed (Figure 9.10). For example, two distinct distribution patterns of lipids were observed in ginkgo leaf. Most lipids were detected in the mesophyll (Figure 9.10), whereas others were only localized in secretory cavities

(Figure 9.9). Phospholipids are more than just structural components of membranes; evidence suggests that they can act as second messengers in plant cells[40] and can be rapidly formed and turned over within seconds to minutes in response to a variety of stress treatments via the activation of lipid kinases or phospholipases. Surprisingly, few spatio‐chemical analyses of lipids

311 in plant tissues have been reported. This is partly due to a lack of specific dyes and labelling techniques. Ion images of lipids should improve the understanding of the distinct function of individual lipids (or lipid classes) and have the potential to aid in deciphering the roles of lipids as signaling molecules in plant growth and development, as well as abiotic and biotic stress responses.

9.5 Acknowledgements

B. L. gratefully acknowledges financial support from the National Natural Science

Foundation of China (no. 81773873). P. L. acknowledges support from the 111 Project (no.

B16046). J. V. S. acknowledges support from the Center for Advanced Bioenergy and

Bioproducts Innovation funded by the U.S. Department of Energy. E. K. N. acknowledges support from the U.S. National Science Foundation Graduate Research Fellowship Program and the Springborn Fellowship. We also thank Weiwei Tang for his assistance in creating histogram graph. The content is solely the responsibility of the authors and does not necessarily represent the official views of the funding agencies.

312

9.6 Figures and Tables

Figure 9.1 Representative single‐pixel matrix‐assisted laser desorption/ionization (MALDI) and laser desorption/ionization (LDI) Fourier‐transform ion cyclotron resonance mass spectra acquired from a cross section of ginkgo leaf. (a, b, c, and e) Positive (POS) mode MALDI mass spectrum of flavonoids obtained from upper epidermis. (d) Positive mode MALDI mass

313

Figure 9.1 (cont.) spectrum of lipids and chlorophylls obtained from mesophyll. (f) Negative

(NEG) mode LDI mass spectrum of ginkgolic acids and cardanols obtained from secretory cavity. (g) Negative mode LDI mass spectrum of flavonoid aglycones obtained from upper epidermis. Identified compounds are labelled with measured mass.

314

Figure 9.2 Representative single-pixel MALDI ToF mass spectra acquired from the cross section of the ginkgo leaf spray-coated with (A) DHB, (B) CHCA, and (C) 9-AA. Identified metabolites are labeled with a red asterisk.

315

Figure 9.3 Representative single-pixel negative mode LDI-ToF mass spectra acquired from a cross section of the ginkgo leaf. Negative mode MALDI mass spectra of (A) flavonoid aglycones and (B) biflavonoids obtained from upper epidermis. (C) Negative mode LDI mass spectrum of ginkgolic acids and cardanols obtained from secretory cavity. Identified specific metabolites are labeled with a red asterisk.

316

Figure 9.4 Representative single-pixel positive mode MALDI-ToF mass spectra acquired from a cross section of the ginkgo leaf (A) spray-coated with DHB and (B) sublimed with DHB.

Identified metabolites are labeled with a red asterisk.

317

Figure 9.5 Optical images of Ginkgo biloba L. leaf, a leaf cross section, and a close‐up view of a cross section of the intervein region and secretory cavity.

318

Figure 9.6 Positive mode MALDI images of selected flavonoid ions, including aglycones

(m/z 271.0601–317.0655), biflavonoids (m/z 539.0973–583.1235), glycosides (m/z 617.1477–

795.1745), and biginkgosides (m/z 1,519.3537–1,551.3435) in ginkgo leaf. Ion images were recorded with a step size of 50 μm. The mass accuracy was less than 2 ppm, and a bin width of m/z = ±5 ppm was used for image generation. Glc: glucoside/glucosyl moiety; K: kaempferol;

Rha: rhamnoside/rhamnosyl moiety; Q: quercetin.

319

Figure 9.7 Negative mode LDI images of ginkgolic acid (GA) ions (m/z 319.2279–373.2748) and cardanol ions (m/z 275.2380–329.2850) in ginkgo leaf. Ion images were recorded with a scanning step size of 50 μm. The mass accuracy was better than 2 ppm, and a bin width of m/z = ±5 ppm was used for image generation. The secretory cavities in the leaf are marked with a red circle.

320

Figure 9.8 Positive mode MALDI images of selected other ions, including saccharides

(m/z 219.0266 and 381.0794), ginkgolides (m/z 447.1052 and 463.1001), an unknown compound

(m/z 742.4805), phosphatidylcholines (PCs) (m/z 796.5253–824.5566), and a fragment ion of chlorophyll a (m/z871.5732) in ginkgo leaf. Ion images were recorded with a scanning step size of 50 μm. The mass accuracy was better than 2 ppm, and a bin width of m/z = ±5 ppm was used for image generation.

321

Figure 9.9 Positive mode MALDI images of unknown ions observed in the secretory cavities of the ginkgo leaf. Ion images were recorded with a scanning step size of 50 μm, and a bin width of m/z = ± 5 ppm was used for image generation.

322

Figure 9.10 Positive mode MALDI images of ginkgolide ions, including ginkgolide A

(m/z 447.1052) and ginkgolide isomers B/J/M/K (m/z 463.1001) in ginkgo root, stem, and leaf.

Regions selected for mass spectrometry imaging measurements are marked with white dashed line. Ion images were recorded with a step size of 50 μm. The mass accuracy was better than

2 ppm, and a bin width of m/z = ±5 ppm was used for image generation.

323

Figure 9.11 Comparison of nine flavonoids in ginkgo leaf between the upper and lower epidermis quantified using liquid chromatography tandem mass chromatography in multiple reaction monitoring mode. Inset displays dissection locations. Error bars indicate the standard deviations of biological triplicates (n = 3). Significant differences were determined using a one‐tailed Student's t‐test. Significance levels: P < 0.05.

324

Figure 9.12 Positive mode MALDI images of unknown ions observed in the secretory cavities of the ginkgo leaf. Ion images were recorded with a scanning step size of 50 μm, and a bin width of m/z = ± 5 ppm was used for image generation.

325

Figure 9.13 Change of flavonoids in ginkgo leaves in response to UV-B radiation quantified using LC-MS/MS in MRM mode. Error bars indicate the standard deviations of biological triplicates (n = 6). Significant differences were determined using a one-tailed Student’s t-test.

Significance levels: *p < 0.05 and **p < 0.01.

326

MRM Rt Assigned Identity MW Transition (min) Isorhamnetin 3-O-rhamnosyl-rhamnosyl- 770.2 769.0>315.0 2.70 glucoside Quercetin 3-O-rhamnosyl-glucoside 610.2 609.0>301.0 2.87 Kaempferol 3-O-rhamnosyl-glucoside 594.2 593.1>285.0 3.07 Isorhamnetin 3-O-rhamnosyl-glucoside 624.2 623.3>315.0 3.12 Kaempferol 286.1 285.0>133.2 4.07 Amentoflavone 538.1 537.0>375.0 5.04 Bilobetin/Sequoiaflavone 552.1 551.2>519.0 5.51 Ginkgetin/Isoginkgetin 566.1 565.0>389.0 6.55 Sciadopitysin 580.1 579.1>547.0 7.63

Table 9.1 MRM transitions of target flavonoids. The identity of target analytes was verified by ESI MS/MS spectra

327

Identified Molecular Exact Measured Error Measured Error Measured Error MS2 Metabolites Formula Mass (u) [M+H]+(u) (ppm) [M+Na]+(u) (ppm) [M+K]+(u) (ppm)

Monosaccharide C6H12O6 180.0634 ‒ ‒ 203.0524 1.04 219.0265 0.22 ‒

Disaccharide C12H22O11 342.1162 ‒ ‒ 365.1056 0.45 381.0795 0.33 ‒

Ginkgolide A C20H24O9 408.1420 ‒ ‒ ‒ ‒ 447.1057 1.13 ‒

Ginkgolide B/J/M/K C20H24O10 424.1370 ‒ ‒ ‒ ‒ 463.1007 1.28 ‒

PC (25:0 (CHO)) C33H64NO9P 649.4319 650.4398 1.00 ‒ ‒ 688.3961 1.55 Yes

PC (25:0(COOH)) C33H64NO10P 665.4268 666.4353 1.86 ‒ ‒ 704.3906 0.93 Yes

PS (P-29:0) C35H68NO9P 677.4632 678.4712 1.11 ‒ ‒ 716.4273 1.35 ‒

PS (P-30:1) C36H68NO9P 689.4632 ‒ ‒ ‒ ‒ 728.4265 0.23 ‒

PA (36:4) C39H69O8P 696.4730 ‒ ‒ ‒ ‒ 735.4363 0.18 ‒

PA (36:3) C39H71O8P 698.4887 ‒ ‒ ‒ ‒ 737.4517 -0.16 ‒

PA (36:2) C39H73O8P 700.5043 ‒ ‒ ‒ ‒ 739.4685 1.40 ‒

PS (P-31:1) C37H70NO9P 703.4788 704.4873 1.71 ‒ ‒ ‒ ‒ ‒

PC (34:3) C42H78NO8P 755.5465 756.5545 0.95 ‒ ‒ 794.5100 0.42 ‒

PC (34:2) C42H80NO8P 757.5622 758.5701 0.88 ‒ ‒ 796.5253 -0.02 Yes

PC (34:1) C42H82NO8P 759.5778 760.5848 -0.40 ‒ ‒ 798.5421 1.42 Yes

PC (36:4) C44H80NO8P 781.5622 782.5701 0.85 ‒ ‒ 820.5250 -0.38 Yes

PC (36:3) C44H82NO8P 783.5778 784.5864 1.68 ‒ ‒ 822.5415 0.65 Yes

PC (36:2) C44H84NO8P 785.5935 786.6015 0.97 ‒ ‒ 824.5580 1.68 Yes

Pheophorbide A C35H36N4O5 592.2686 593.2742 -2.78 ‒ ‒ ‒ ‒ ‒

Pheophytin A C55H74N4O5 870.5659 871.5717 -1.72 893.5564 1.41 909.5298 0.79 ‒

Chlorophyll A C55H72MgN4O5 892.5353 ‒ ‒ ‒ ‒ 931.4977 -0.83 ‒

Table 9.2 Saccharides, phospholipids, and chlorophylls assigned in G. biloba leaf tissues by

positive mode MALDI FT-ICR MS. Metabolites were putatively identified based on high mass

accuracy of full scan MS data and/or MS/MS.

328

Molecular Exact Measured Error Measured Error Measured Error Identified Metabolites MS2 Formula Mass (u) [M+H]+(u) (ppm) [M+Na]+(u) (ppm) [M+K]+(u) (ppm)

Apigenin C15H10O5 270.0528 271.0597 -1.48 ̶ ̶ ̶ ̶ ̶

Kaempferol/Luteolin C15H10O6 286.0477 287.0545 -1.80 ̶ ̶ 325.0103 -1.84 ̶

Catechin/Epicatechin C15H14O6 290.0790 291.0862 -0.40 ̶ ̶ ̶ ̶ ̶

Quercetin C15H10O7 302.0427 303.0494 -1.75 ̶ ̶ ̶ ̶ Yes

Tamarixetin/Isorhamnetin C16H12O7 316.0583 317.0650 -1.83 ̶ ̶ ̶ ̶ Yes

Myricetin C15H10O8 318.0376 319.0445 -1.09 ̶ ̶ ̶ ̶ ̶

Amentoflavone C30H18O10 538.0900 539.0967 -1.07 ̶ ̶ ̶ ̶ ̶

Bilobetin/Sequoiaflavone C31H20O10 552.1057 553.1124 -0.95 ̶ ̶ ̶ ̶ ̶

Ginkgetin/Isoginkgetin C32H22O10 566.1213 567.1285 -0.13 589.1096 -1.56 605.0843 -0.26 ̶

Sciadopitysin C33H24O10 580.1370 581.1435 -1.25 603.1255 -1.11 619.1001 -0.01 ̶

5′-Methyoxy bilobetin C32H22O11 582.1162 583.1224 -1.87 ̶ ̶ ̶ ̶ ̶

Quercetin 3-O- C21H20O11 448.1006 449.1076 -0.54 471.089 -1.67 487.0628 -1.89 ̶ rhamnoside Quercetin 3-O-glucoside C21H20O12 464.0955 465.1031 0.74 487.0839 -1.64 ̶ ̶ ̶

Isorhamnetin 3-O- C22H22O12 478.1111 479.1181 -0.64 ̶ ̶ ̶ ̶ ̶ glucoside Kaempferol 3-O- C27H30O15 594.1585 ̶ ̶ 617.1466 -1.78 633.121 -1.00 Yes rhamnosyl-glucoside/3-O- glucosyl-rhamnoside Quercetin 3-O-rhamnosyl- C27H30O16 610.1534 ̶ ̶ 633.1416 -1.60 649.1153 -1.92 Yes glucoside/3-O-glucosyl- rhamnoside Isorhamnetin 3-O- C28H32O16 624.1690 ̶ ̶ 647.1571 -1.79 663.1317 -0.75 Yes rhamnosyl-glucoside Methoxymyricetin 3-O- C28H32O17 640.1640 663.1537 0.79 679.1271 -1.93 Yes rhamnosyl-glucoside Kaempferol 3-O-(p- C36H36O17 740.1953 ̶ ̶ 763.1845 0.03 779.1583 -0.15 Yes coumaroylglucosyl- rhamnoside) Kaempferol 3-O- C33H40O19 740.2164 ̶ ̶ 763.2057 0.12 779.1806 1.36 ̶ rhamnosyl-rhamnosyl- glucoside Quercetin 3-O-(p- C36H36O18 756.1902 ̶ ̶ 779.1779 -1.91 795.1533 -0.04 ̶ coumaroylglucosyl- rhamnoside) Quercetin 3-O-(2-O,6-O- C33H40O20 756.2113 ̶ ̶ ̶ ̶ 795.1755 1.31 ̶ bis(rhamnosyl)-glucoside) Isorhamnetin 3-O- C34H42O20 770.2270 ̶ ̶ 793.2166 0.54 809.1906 0.61 ̶ rhamnosyl-rhamnosyl- glucoside

Biginkgoside A/C/D/G/H C72H72O34 1480.3905 ̶ ̶ ̶ ̶ 1519.3557 1.34 Yes

Biginkgoside B/E/I C72H72O36 1512.3803 ̶ ̶ ̶ ̶ 1551.3435 0.001 Yes

Biginkgoside F C72H72O35 1496.3854 ̶ ̶ ̶ ̶ 1535.3476 -0.67 Yes

Table 9.3 Flavonoids assigned in G. biloba leaf tissues by positive mode MALDI FT-ICR MS.

Metabolites were putatively identified based on high mass accuracy of full scan MS data and/or

MS/MS.

329

Molecular Exact Measured Error Identified Metabolites MS2 Formula Mass (u) [M-H]- (u) (ppm) Yes Ginkgolic acid C13:0 C20H32O3 320.2351 319.2285 1.98 Yes Ginkgolic acid C15:2 C22H32O3 344.2351 343.2285 1.84

Ginkgolic acid C15:1 C22H34O3 346.2508 345.2439 1.10 Yes Yes Ginkgolic acid C15:0 C22H36O3 348.2664 347.2598 1.82 Yes Ginkgolic acid C17:3 C24H34O3 370.2508 369.2442 1.85 Yes Ginkgolic acid C17:2 C24H36O3 372.2664 371.2598 1.70 Yes Ginkgolic acid C17:1 C24H38O3 374.2821 373.2754 1.56 Yes Cardanol C13:0 C19H32O 276.2453 275.2385 1.68 Yes Cardanol C15:1 C21H34O 302.2610 301.2541 1.37 ̶ Cardanol C15:0 C21H36O 304.2766 303.2699 1.85 Yes Cardanol C17:1 C23H38O 330.2923 329.2853 0.95 ̶ Kaempferol/Luteolin C15H10O6 286.0477 285.0408 1.18 ̶ Quercetin C15H10O7 302.0427 301.0357 1.07 ̶ Tamarixetin/Isorhamnetin C16H12O7 316.0583 315.0515 1.50 Yes Amentoflavone C30H18O10 538.0900 537.0827 -0.04 Yes Bilobetin/Sequoiaflavone C31H20O10 552.1057 551.0989 0.96 Yes Ginkgetin/Isoginkgetin C32H22O10 566.1213 565.1149 1.55 Yes Sciadopitysin C33H24O10 580.1370 579.1302 0.91 Yes 5′-Methyoxy bilobetin C32H22O11 582.1162 581.1099 1.65 Quercetin 3-O- ̶ rhamnosyl-glucoside/3- C27H30O16 610.1534 609.1469 1.29 O-glucosyl-rhamnoside Kaempferol 3-O-(p- ̶ coumaroylglucosyl- C36H36O17 740.1953 739.1882 0.30 rhamnoside) Quercetin 3-O-(p- ̶ coumaroylglucosyl- C36H36O18 756.1902 755.1837 1.07 rhamnoside)

Table 9.4 Ginkgolic acids, cardanols, and lavonoids assigned in G. biloba leaf tissues by negative mode MALDI FT-ICR MS. Metabolites were putatively identified based on high mass accuracy of full scan MS data and/or MS/MS.

330

Identified Molecular Exact Measured Error Measured Error Measured Error MS2 Metabolites Formula Mass (u) [M+H]+(u) (ppm) [M+Na]+(u) (ppm) [M+K]+(u) (ppm)

Monosaccharide C6H12O6 180.0634 ‒ ‒ 203.0524 1.04 219.0265 0.22 ‒

Disaccharide C12H22O11 342.1162 ‒ ‒ 365.1056 0.45 381.0795 0.33 ‒

Ginkgolide A C20H24O9 408.1420 ‒ ‒ ‒ ‒ 447.1057 1.13 ‒

Ginkgolide B/J/M/K C20H24O10 424.1370 ‒ ‒ ‒ ‒ 463.1007 1.28 ‒

PC (25:0 (CHO)) C33H64NO9P 649.4319 650.4398 1.00 ‒ ‒ 688.3961 1.55 Yes

PC (25:0(COOH)) C33H64NO10P 665.4268 666.4353 1.86 ‒ ‒ 704.3906 0.93 Yes

PS (P-29:0) C35H68NO9P 677.4632 678.4712 1.11 ‒ ‒ 716.4273 1.35 ‒

PS (P-30:1) C36H68NO9P 689.4632 ‒ ‒ ‒ ‒ 728.4265 0.23 ‒

PA (36:4) C39H69O8P 696.4730 ‒ ‒ ‒ ‒ 735.4363 0.18 ‒

PA (36:3) C39H71O8P 698.4887 ‒ ‒ ‒ ‒ 737.4517 -0.16 ‒

PA (36:2) C39H73O8P 700.5043 ‒ ‒ ‒ ‒ 739.4685 1.40 ‒

PS (P-31:1) C37H70NO9P 703.4788 704.4873 1.71 ‒ ‒ ‒ ‒ ‒

PC (34:3) C42H78NO8P 755.5465 756.5545 0.95 ‒ ‒ 794.5100 0.42 ‒

PC (34:2) C42H80NO8P 757.5622 758.5701 0.88 ‒ ‒ 796.5253 -0.02 Yes

PC (34:1) C42H82NO8P 759.5778 760.5848 -0.40 ‒ ‒ 798.5421 1.42 Yes

PC (36:4) C44H80NO8P 781.5622 782.5701 0.85 ‒ ‒ 820.5250 -0.38 Yes

PC (36:3) C44H82NO8P 783.5778 784.5864 1.68 ‒ ‒ 822.5415 0.65 Yes

PC (36:2) C44H84NO8P 785.5935 786.6015 0.97 ‒ ‒ 824.5580 1.68 Yes

Pheophorbide A C35H36N4O5 592.2686 593.2742 -2.78 ‒ ‒ ‒ ‒ ‒

Pheophytin A C55H74N4O5 870.5659 871.5717 -1.72 893.5564 1.41 909.5298 0.79 ‒

Chlorophyll A C55H72MgN4O5 892.5353 ‒ ‒ ‒ ‒ 931.4977 -0.83 ‒

Table 9.5 Saccharides, phospholipids, and chlorophylls assigned in G. biloba leaf tissues by

positive mode MALDI FT-ICR MS. Metabolites were putatively identified based on high mass

accuracy of full scan MS data and/or MS/MS.

331

9.7 References

(1) B. Singh, P. Kaur, Gopichand, R. D. Singh, P. S. Ahuja, Fitoterapia 2008, 79, 401-418.

(2) M. Hemshekhar, M. Sebastin Santhosh, K. Kemparaju, K. S. Girish, Basic & Clinical

Pharmacology & Toxicology 2012, 110, 122-132.

(3) T. A. van Beek, Journal of Chromatography A 2002, 967, 21-55.

(4) B. J. Diamond, S. C. Shiflett, N. Feiwel, R. J. Matheis, O. Noskin, J. A. Richards, N. E.

Schoenberger, Archives of physical medicine and rehabilitation 2000, 81, 668-678.

(5) W. Gong, C. Chen, C. Dobeš, C.-X. Fu, M. A. Koch, Molecular Phylogenetics and

Evolution 2008, 48, 1094-1105.

(6) R. B. Goldberg, Science 1988, 240, 1460-1467.

(7) B. A. Boughton, D. Thinagaran, D. Sarabia, A. Bacic, U. Roessner, Phytochemistry

Reviews 2016, 15, 445-488.

(8) A. D. Feenstra, L. E. Alexander, Z. Song, A. R. Korte, M. Yandeau-Nelson, B. J.

Nikolau, Y.-J. Lee, Plant physiology 2017, pp. 00652.02017.

(9) S. Kaspar, M. Peukert, A. Svatos, A. Matros, H.-P. Mock, PROTEOMICS 2011, 11,

1840-1850.

(10) Y. J. Lee, D. C. Perdian, Z. Song, E. S. Yeung, B. J. Nikolau, The Plant Journal 2012,

70, 81-95.

(11) Z. Li, P. W. Bohn, J. V. Sweedler, Bioresource Technology 2010, 101, 5578-5585.

(12) D. S. Cornett, M. L. Reyzer, P. Chaurand, R. M. Caprioli, Nature Methods 2007, 4, 828.

(13) A. Brunelle, D. Touboul, O. Laprévote, Journal of Mass Spectrometry 2005, 40, 985-999.

(14) Z. Takáts, J. M. Wiseman, B. Gologan, R. G. Cooks, Science 2004, 306, 471-473.

(15) D. Etalo, R. C. De Vos, M. H. Joosten, R. Hall, Plant physiology 2015, pp. 01176.02015.

332

(16) P. Nemes, A. Vertes, Analytical Chemistry 2007, 79, 8098-8106.

(17) M. Kompauer, S. Heiles, B. Spengler, Nature Methods 2016, 14, 90.

(18) Y. Dong, B. Li, S. Malitsky, I. Rogachev, A. Aharoni, F. Kaftan, A. Svatoš, P.

Franceschi, Frontiers in Plant Science 2016, 7.

(19) B. Li, T. J. Comi, T. Si, S. J. B. Dunham, J. V. Sweedler, Journal of Mass Spectrometry

2016, 51, 1030-1035.

(20) E. J. Lanni, S. J. B. Dunham, P. Nemes, S. S. Rubakhin, J. V. Sweedler, Journal of The

American Society for Mass Spectrometry 2014, 25, 1897-1907.

(21) T. Yamagaki, T. Watanabe, M. Tanaka, K. Sugahara, Journal of The American Society

for Mass Spectrometry 2014, 25, 88-94.

(22) P. Česla, L. Blomberg, M. Hamberg, P. Jandera, Journal of Chromatography A 2006,

1115, 253-259.

(23) B. Li, S. J. B. Dunham, Y. Dong, S. Yoon, M. Zeng, J. V. Sweedler, Trends in Food

Science & Technology 2016, 47, 50-63.

(24) S. A. Schwartz, M. L. Reyzer, R. M. Caprioli, Journal of Mass Spectrometry 2003, 38,

699-708.

(25) G. Hrazdina, in Phenolic Metabolism in Plants (Eds.: H. A. Stafford, R. K. Ibrahim),

Springer US, Boston, MA, 1992, pp. 1-23.

(26) P. CASATI, V. WALBOT, Plant, Cell & Environment 2005, 28, 788-799.

(27) G. Liang, Y. Guo, B. G. Rector, Journal of Economic Entomology 2003, 96, 909-913.

(28) G.-L. Ma, J. Xiong, G.-X. Yang, L.-L. Pan, C.-L. Hu, W. Wang, H. Fan, Q.-H. Zhao, H.-

Y. Zhang, J.-F. Hu, Journal of Natural Products 2016, 79, 1354-1364.

(29) K. Gorzolka, H. Bednarz, K. Niehaus, Planta 2014, 239, 1321-1335.

333

(30) B. Li, C. Knudsen, N. K. Hansen, K. Jørgensen, R. Kannangara, S. Bak, A. Takos, F.

Rook, S. H. Hansen, B. L. Møller, C. Janfelt, N. Bjarnholt, The Plant Journal 2013, 74,

1059-1071.

(31) I. Kubo, K.-i. Nihei, K. Tsujimoto, Journal of Agricultural and Food Chemistry 2003, 51,

7624-7628.

(32) A. Fahn, R. F. Evert, American Journal of Botany 1974, 61, 1-14.

(33) A. FAHN, D. M. JOEL, Annals of Botany 1980, 46, 225-233.

(34) T. B. Andersen, K. A. Martinez-Swatson, S. A. Rasmussen, B. A. Boughton, K.

Jørgensen, J. Andersen-Ranberg, N. Nyberg, S. B. Christensen, H. T. Simonsen, Plant

physiology 2017, 174, 56-72.

(35) B. M. Lange, Annual Review of Plant Biology 2015, 66, 139-159.

(36) S. N. Jackson, H.-Y. J. Wang, A. S. Woods, M. Ugarov, T. Egan, J. A. Schultz, Journal

of The American Society for Mass Spectrometry 2005, 16, 133-138.

(37) A. Cartayrade, E. Neau, C. Sohier, J. Balz, J. Carde, J. Walter, Plant Physiology and

Biochemistry (France) 1997.

(38) X. Lu, H. Yang, X. Liu, Q. Shen, N. Wang, L.-w. Qi, P. Li, Frontiers in Plant Science

2017, 8.

(39) N. Tocci, M. Gaid, F. Kaftan, A. K. Belkheir, I. Belhadj, B. Liu, A. Svatoš, R. Hänsch,

G. Pasqua, L. Beerhues, New Phytologist 2018, 217, 1099-1112.

(40) H. J. G. Meijer, T. Munnik, Annual Review of Plant Biology 2003, 54, 265-306.

334