Determination of Optimal Electrospray Parameters for Lipidomics in Infrared Matrix Assisted Laser Desorption Electrospray Ionization (IR-MALDESI) Mass Spectrometry Imaging

Abstract

Infrared matrix-assisted laser desorption ionization (IR-MALDESI) is an ambient mass spectrometry imaging (MSI) technique that relies on electrospray ionization (ESI) for ion generation of desorbed neutrals. Although many mechanisms in IR-MALDESI have been studied in depth, there has not yet been a comprehensive study of how the ESI parameters change the profiles of tissue specific lipids. Acetonitrile (ACN):water and methanol (MeOH):water solvent systems and compositions were varied across a series of applied ESI voltages during IR-MALDESI analysis of rat liver tissue. Twelve-minute gradients were run from 5–95% organic solvent in both positive and negative polarities across eleven voltages between 2.25 kV—4.5 kV. These experiments informed longer gradients (25—30 minutes) across shorter solvent gradient ranges with fewer voltages. Optimal ESI parameters for lipidomics were determined by the number and abundance of detected lipids and the relative proportion of background ions. In positive polarity the best solvent composition was 60—75% ACN/40—25% H₂O with 0.2% formic acid at 3.2 kV applied voltage. The best parameters for negative polarity analysis are 45—55% ACN/55—45% H₂O with 1 mM of acetic acid for voltages between 2.25—3.2 kV. Using these defined parameters IR-MALDESI positive polarity lipidomics studies can increase lipid abundances 3-fold, with 15% greater coverage, while an abundance increase of 1.5-fold and 10% more coverage can be achieved relative to commonly used parameters in negative polarity.

INTRODUCTION

Since its characterization as an ionization technique [1,2] and subsequent discovery as viable for biomolecules [3], electrospray ionization (ESI) with mass spectrometry (MS) analysis has transformed the field of bioanalysis. In lipidomics, researchers have sought to separate lipids from other small metabolites and lipids using two major ESI MS approaches: “shotgun” lipidomics or liquid chromatography (LC) separation prior to ESI. [4,5] In “shotgun” lipidomics the entire sample is analyzed without any separation prior to MS analysis. In “shotgun” lipidomics, many studies operate using an “intrasource” separation strategy to affect the categories and classes of lipids analyzed and quantified, which is achieved by specific sample preparation, addition of internal standards and changing ESI parameters.[5] In LC-MS, lipids are separated over time, usually based on headgroup polarity, although they can also be separated by length and degree of saturation.[6] Paired with separation methods in either type of study, tandem mass spectrometry (MS/MS) fragmentation of lipid precursors and subsequent identification, provides the highest levels of separation and identification of lipids across the major lipid categories and classes. In both types of analysis ESI variables play an outsized role in the identified lipids with parameters such as applied voltage, ionization mode, solvent and solvent composition demonstrated to affect the ionization efficiencies of different types of lipids.[5,7] While MS/MS is necessary for confident identification of lipids, often lipids can be separated and putatively assigned using high RP instruments that distinguish between ions with mass-to-charge (m/z) differences of a few parts per million (ppm) or less; although it should be noted that high RP cannot separate between isomeric species and sometimes cannot resolve isobaric species.

ESI-based mass spectrometry imaging (MSI) methods of endogenous lipids inherently rely on ESI parameters for “intrasource”-like separation between lipids, as well as high resolution accurate mass (HRAM) analysis for lipid identification. Further identification using MS/MS in MSI has been demonstrated before but is often impractical on a large scale and only performed in follow up experiments on lipids with important spatial distributions.[8,9] In desorption electrospray ionization (DESI), the effects of solvent compositions on metabolites extracted from the surface of samples have been modelled to develop methodologies to increase desired ion abundances and limit background ions in the interrogated m/z range.[10] In fact, most MSI techniques have been well characterized regarding their extraction, ionization, and analysis for different types of molecules. Matrix assisted laser desorption ionization (MALDI) is arguably the most extensively characterized, due in large part to its widespread use in MSI. [11] Despite difficulties in analyzing small metabolites, a great focus has been placed on MALDI imaging of lipids, from the appropriate laser to the matrix with fewest low m/z interferents. [12,13]

Infrared matrix-assisted laser desorption electrospray ionization (IR-MALDESI) was invented to combine many of the benefits of associated with MALDI and ESI. IR-MALDESI works by combining laser desorption of neutrals with subsequent ionization by ESI. In a typical IR-MALDESI experiment, tissue samples are placed on a cooled translation stage in a humidity-controlled enclosure. An ice layer is formed on top of the sample for optimal plume dynamics and demonstrably increased ion analysis efficiency.[14,15] A mid-IR laser with wavelength of 2.94 μm, in resonance with the O—H stretching mode of the endogenous water and exogenous ice matrix, desorbs a plume of neutral molecules from the sample, which is the “extraction” mechanism that extracts diverse chemical classes from the sample. Those neutral molecules partition into droplets in an orthogonal electrospray plume and are ionized in an ESI-like manner before being analyzed and separated using the high RP (140,000 at 200 m/z) of an Orbitrap MS; analogous to other “shotgun” –omics strategies, with the added benefits of spatially resolved analysis of metabolites in diverse sample types. [16–18] In this manner IR-MALDESI is an ambient ionization method with minimal sample preparation and ionizes molecules using ESI, while offering the ability to perform absolute quantification of metabolites due to a reproducible volume of desorbed tissue and stable isotope labeled internal standard workflows.

To date, there has not been a comprehensive study of the effects that electrospray solvent composition, ionizing additives, applied voltages and ionization mode have on lipid profiles during IR-MALDESI analyses. A previous study characterized a few lipid species, background ions, and a drug as one part of a larger project.[19] In this paper we report the effects of systematically changing ESI parameters on the abundance of tissue specific lipids in rat liver tissues. Understanding the ionization biases in IR-MALDESI will enable greater ionization specificity without losing the benefits inherent to it as an ambient technique.

MATERIALS AND METHODS

Rat Liver Tissue Preparation

The livers were frozen in an isopentane/dry ice bath and stored at −80°C. The tissue was removed from −80°C, placed on a sample disk with optimal cutting temperature (OCT) mounting compound, and solidified at −20°C. Sections were made using a Leica CM1950 cryostat (Buffalo, Grove, IL, USA) and the 25 μm sections were thaw-mounted on pre-cleaned microscope slides.[20]

IR-MALDESI Imaging Analysis of Rat Livers

IR-MALDESI has been described in detail previously (Supplemental Information Figures 1, 2).[21] Briefly, the home-built source consists of a humidity-controlled enclosure, a Peltier-cooled translation stage, a mid-IR laser and a Q Exactive Plus mass spectrometer (Thermo Fisher Scientific, San Jose, CA, USA). Samples were placed on the cooled stage, and the enclosure was purged with nitrogen gas (ARC3 Gases, Raleigh, NC, USA) to a relative humidity of 10% or less. The stage was held at −8°C for several minutes before exposure to ambient humidity and forming an ice layer on top of the tissue. The mid-IR (2.94 μm wavelength) laser Opolette 2731/3034 (Opotek, Carlsbad, CA, USA) was fired at the sample with a single pulse per acquisition. The injection time was fixed at 25 ms, the minimum IT to capture all ions generated by the laser pulse and maximize the abundance of target ions.[22] Resolving power (RP) was set to 140,000 FWHM at m/z 200. Laser spot size was ~ 150 μm and step size was 100 μm. Raw data files from each imaging experiment were converted to mzML using MSConvert [23,24] and then to imzML using imzMLConverter [25] before being loaded into MSiReader v1.01 [26,27] for image analysis and data consolidation.

Liquid Chromatography Gradient Methods

An Eksigent ekspert nanoLC 415 (AB Sciex, Framingham, MA, USA) was used to create solvent gradients for both water:acetonitrile (ACN) and water:methanol (MeOH) (all: Optima^® LC/MS Grade, Fisher Chemical, Fairlawn, NJ, USA). They were run through a capillary without a column at a flowrate of 2 μL/min. Each solvent, MeOH, ACN, and water, contained 0.2% formic acid for positive polarity measurements. The gradients were run from 5—95% ACN or MeOH linearly in 12 minutes with 1 minute at both ends of the gradient (14 minutes total). The voltages were changed between each of the 11 values shown in Figure 1 over the course of the entire gradient method. The same experiments were performed in negative polarity, except 1 mM acetic acid was used instead of 0.2% formic acid.[28,29] Once the gradient solvent composition range had been determined longer 25-minute gradients were run over the new shorter solvent composition ranges.

Figure 1.

ACN, MeOH gradients were run from 5—95% with voltages from 2.25 to 4.5 kV in both positive and negative polarities

RESULTS AND DISCUSSION

The electrospray onset voltage was measured for the extreme points of the gradient (5% organic and 5% aqueous), and used to define the boundary conditions to interrogate. The onset voltages for MeOH and ACN solvent systems were measured in positive and negative polarities. They differed by as much as 200 V between replicate experiments. In general, measured onset voltages for the IR-MALDESI system are higher (100—300 V) than the calculated values for the current IR-MALDESI setup including capillary specifications and distance between electrodes.[30,31] Equation 1 shows the relationship of onset voltage and how it is related to solvent composition (surface tension) (Figure 2).

Figure 2.

Onset voltages predictions for IR-MALDESI (black dots) change with solvent composition (red dots). Measured onset voltages in IR-MALDESI in positive and negative polarities at both 5:95 and 95:5 water:solvent endpoints are shown as stars in red (positive) and blue (negative)

$$V_{on}=2x{10}^5\sqrt{\gamma r_c}\ln \left(\frac{4d}{r_c}\right)\stackrel{\text{⬚}}{\Rightarrow }{V}_{on}\propto \sqrt{\gamma }$$ (1)

Once the onset voltages were found, short gradients (12 minutes) were run from 5—95% organic solvent. In positive mode, 85 METASPACE assignable tissue specific lipids were classified using LIPIDMAPS and tracked across the gradients and voltages (Supplemental Information). These lipids were a subset of the ions used in previous studies for evaluation of IR-MALDESI repeatability.[32] In addition to comparison of spectra with ESI background spectra (Supplemental Figure 3), steps were taken to ensure their identity as tissue specific ions by way of METASPACE (metaspace2020.eu) [33] and METLIN [34,35] metabolite identifications, followed by LIPID MAPS categorization [36–38]. Identification was limited to molecular formula as isomers were not resolved. Common adducts in MALDESI (i.e. protonation, water loss, and deprotonation) were given higher preference in identifications of lipids, but other adducts (sodium, potassium) could have been present as well and introduce a small source of potential error in identifications. Sodium cationization would be the major form of ionization if the ESI was doped. Such an experiment has been previously demonstrated using ESI doped with silver nitrate.[39] In positive mode, six of the eight major lipid categories were covered with no saccharolipids or polyketides in the selected lipids list, while in negative mode the seven relevant major categories were observed (Supplemental Information). The average abundance for peak detection was on the order of 10e3 AU and the average abundance of most observed lipids was on the order of 10e5. Figure 3 illustrates that the most abundant responses for the 85 assignable tissue specific lipids for each solvent system were found at lower voltages, closer to onset voltages. Figure 4 informed the next round of experiments by illustrating that when MeOH content was 45—75% and ACN content was 35—75%. This informed the 25-minute gradients run over a smaller solvent composition range, for fewer voltages (Supplemental Information Table 1).

Figure 3.

(a) Total ion current (TIC) of 85 tissue specific lipids plotted across 5–95% ACN gradients for voltages from 2.25 kV to 4.5 kV (blue) and percent coverage of those lipids (green). Percent coverage is defined as number of lipids detected/85 total assignable lipids. (b) TIC of tissue specific lipids plotted as a ratio of the Total TIC across voltages to determine which resulted in greater relative abundance of the lipids. (c,d) Panels are for MeOH

Figure 4.

Relative abundances of 85 tissue specific lipids across the gradients of (a) ACN and (b) MeOH. They are shown from low to high m/z from left to right across each image. Each column contains abundance data that is relatively scaled per lipid and plotted linearly from black (low) to red (middle) to yellow (high) to determine which gradient range was best for each lipid.

In Figure 5 the best summed lipid abundance and lipid coverage, of the 85 tissue specific lipids was found in the 2.75–3.2 kV voltage range and between the 55–75% ACN. The best combination is 2.75–3.2 kV and 65–75% for MeOH. These experiments clarify that for tissue-related lipids, MALDESI experiments need to operate with ESI solutions with higher organic solvent concentrations and at lower voltages than used in previous experiments. Specifically, the best overall is 60–75% ACN at 3.2 kV and for MeOH, 65–75% at 3.0 kV, with ACN a better solvent choice than MeOH.

Figure 5.

(a, d) The TIC of 85 tissue specific lipids is plotted in blue and percent coverage of those lipids is in green. (b, e) The TIC of tissue specific lipids is plotted as a ratio of the whole TIC. (c, f) The lipid abundances are plotted as mentioned in Figure 4, except they are shown from low to high m/z from top to bottom of each image. a, b, c represent the ACN gradients, and d, e, f represent MeOH gradients

After the positive polarity experiments were completed, they were repeated for negative ion mode using the same solvents, except 1 mM acetic acid was used instead of formic acid. Short gradients (12 minutes) were run from 5—95% organic solvent. There were 133 tissue specific lipids interrogated in negative ion mode, which were determined by comparing the overlap in tissue specific lipids between MeOH and ACN gradients. Similar to the 85 lipids in positive mode, putative identifications were made using METASPACE [33] and METLIN [34,35], with LIPID MAPS categorization demonstrating that the seven relevant major lipid categories were covered based on MS1 (Supplemental Information) [36–38]. They were tracked along the gradients and voltages to inform the solvent composition ranges and voltages for further analysis over longer gradients (Supplemental Information). Figure 6 illustrates that when MeOH was 10—55% and voltages were between 3.0—4.5 kV, tissue specific lipids were best analyzed. Low compositions of MeOH in particular resulted in higher ion abundances (Figure 7).

Figure 6.

Relative abundances of 133 tissue specific lipids found in both gradients of ACN (a) and MeOH (b). They are shown from low to high m/z from left to right across each image. Each column contains abundance data. Abundance per lipid is plotted linearly from black (low) to blue (middle) to yellow (high) to determine which gradient range was best per lipid and for the combined group

Figure 7.

(a) Total ion current (TIC) of 133 assignable tissue specific lipids plotted across ACN gradients for voltages from 2.25 kV to 4.5 kV (blue) and percent coverage of those lipids (green). Percent coverage is defined as number of lipids measured divided by 133 assignable lipids. (b) TIC of tissue specific lipids plotted as a ratio of the Total TIC across voltages to determine which resulted in greater relative abundance of the lipids. (c, d) Panels are for MeOH

However, due to the tendency to slowly form droplets at high aqueous compositions, the longer gradients (25 minutes) started at 20% MeOH, instead of 10%. The combinations selected for analysis for ACN were 40—75 % at 2.25—4.0 kV and 20—55% MeOH at 3.0–4.5 kV (Supplemental Information Table 2).

As shown in Figure 8, for the ACN gradients between 40—75% and 2.25—3.8 kV there is not one best voltage/solvent composition, nor is there a consistent gradient range that is best between all voltages. In general, the best lipid abundances were found at relatively low voltages and ACN content (2.25—3.2 kV, 45—55%), or higher relative voltages and ACN content (3.4—3.8 kV, 60—70%). Across the MeOH gradient there were no improvements between voltages or solvent composition. However, at lower MeOH compositions many droplets formed that either resulted in pulsed ESI or were wiped off before they fell onto the tissue. The percent coverage of tissue specific lipids were consistently in the 30—40% range, below that often found in the ACN experiments. Appropriate combinations for negative polarity experiments would be 45—55% ACN and 2.25—3.2 kV, above that voltage from 3.4—3.8 kV the best combination is 60—70%.

Figure 8.

(a, d) The tissue specific lipid abundances are plotted as mentioned in Figure 5, shown from low to high m/z from top to bottom of each image. (b, e) the TIC of 133 assignable tissue specific lipids is plotted in blue and percent coverage of those lipids is in green. (c, f) In c and f the TIC of 133 assignable tissue specific lipids is plotted as a ratio of the whole TIC. A, b, c model the ACN gradients, and d, e, f model MeOH gradients

CONCLUSIONS

The effects of varying IR-MALDESI electrospray parameters for analysis of low m/z (250—1000 m/z) lipids from rat liver tissue were determined in positive and negative polarities. In positive mode, the best solvent compositions for lipid analysis of rat liver tissues were determined to be 60—75% ACN/40—25% H₂O with 0.2% formic acid at 3.2 kV applied voltage. Using these ESI parameters, an average of a 3-fold increase in abundance and a 15% increase in percent coverage of assignable tissue specific lipids relative to previously used ESI parameters (50:50 MeOH:H₂O, 0.2% formic acid, 3—4.2 kV) were achieved. This conclusion was based on absolute ion abundance of 85 assignable tissue specific lipids, the percent coverage of those lipids, stability of signal during measurements and the relative abundance of background ions. These same metrics were applied to determine that the best parameters for negative mode analysis of 133 assignable tissue specific lipids are 45—55% ACN/55—45% H₂O with 1 mM of acetic acid for voltages between 2.25—3.2 kV. These newly defined parameters result in an average 1.5-fold increase in abundance and a 10% increase in percent coverage of selected tissue specific lipids relative to previously used ESI parameters (50:50 MeOH;H₂O, 1 mM acetic acid, 3—4.2 kV). Future lipidomics work using IR-MALDESI are now informed of appropriate ESI parameters to increase the abundances of desired lipids and lipid classes, while limiting background species.