Automated 3D Sampling and Imaging of Uneven Sample Surfaces with LA-REIMS.

The analysis of samples with large height variations remains a challenge for mass spectrometry imaging (MSI), despite many technological advantages. Ambient sampling and ionization MS techniques allow for the molecular analysis of sample surfaces with height variations, but most techniques lack MSI capabilities. We developed a 3D MS scanner for the automated sampling and imaging of a 3D surface with laser-assisted rapid evaporative ionization mass spectrometry (LA-REIMS). The sample is moved automatically with a constant distance between the laser probe and sample surface in the 3D MS Scanner. The topography of the surface was scanned with a laser point distance sensor to define the MS measurement points. MS acquisition was performed with LA-REIMS using a surgical CO2 laser coupled to a qTOF instrument. The topographical scan and MS acquisition can be completed within 1 h using the 3D MS scanner for 300 measurement points on uneven samples with a spatial resolution of 2 mm in the top view, corresponding to 22.04 cm2. Comparison between the automated acquisition with the 3D MS scanner and manual acquisition by hand showed that the automation resulted in increased reproducibility between the measurement points. 3D visualizations of molecular distributions related to structural differences were shown for an apple, a marrowbone, and a human femoral head to demonstrate the imaging feasibility of the system. The developed 3D MS scanner allows for the automated sampling of surfaces with uneven topographies with LA-REIMS, which can be used for the 3D visualization of molecular distributions of these surfaces.

Introduction

Over the past decades, mass spectrometry imaging (MSI) has been increasingly used to investigate biological and biomedical samples. Nevertheless, the surface sampling of large areas with topographical height variations with 3D MSI is very limited. The analysis of samples with large and uneven surfaces has a potentially broad range of applications, including for the analysis of whole fruits or vegetables as well as clinically for analysis of large surface areas on bones and joints. The analysis of these types of surfaces can be of added value in reproducible molecular mapping of samples surfaces to study the molecular changes on the sample surface. For clinical applications, this can be used to improve molecular understanding of processes related to structural changes, which can be applied to tissue classification model building for identification of diseases. In addition, the number of sampling spots, and thus, tissue damage can be reduced by determining the optimal sampling point due to the improved molecular understanding of the sample surface for further applications.

Samples with uneven surfaces are more challenging to analyze with MSI, as mass analyzers, like time-of-flight (TOF) analyzers, are frequently affected by small height variations in the sample resulting in signal intensity variations or peak shifts.[1] For example, a linear TOF is more sensitive to height variation than an orthogonal TOF, as in an orthogonal TOF the ionization source is decoupled from the mass analysis.[2] Height variations of even a few micrometers in the sample can cause defocusing of the laser beam during analysis for MSI technique relying on laser beams, resulting in decreasing sensitivity.[3] Because of these challenges, most MSI techniques require a constant distance between the sample, the ionization source, and the extraction to the mass analyzer. Samples prepared for 2D MSI techniques, such as matrix-assisted laser desorption/ionization (MALDI) and desorption electrospray ionization (DESI), are often sectioned to create planar surfaces for easy analysis.[3−6] Nevertheless, sectioning is not always an option for certain tissue types or surface analyses.

Some techniques have been developed that can compensate for small variations in sample height. These compensation techniques are based on obtaining surface height measurements by, for example, camera image analysis,[7,8] atomic force microscopy,[9−12] shear force microscopy,[13] scanning near-field optical microscopy,[14,15] laser triangulation,[5,16] or confocal distance sensor.[3] The height variation tolerance remains limited to only hundreds of μms and still requires the (biological) samples to be relatively flat for all these compensation techniques. Therefore, the analysis of biological sample surfaces with large height variations in terms of millimeters, for example, bones and joints removed during surgery, is not possible with these developments.

It is important to determine the topography of the surface that needs to be analyzed to achieve MSI of 3D surfaces. This topography will determine the position of the sample, ionization source, and MS inlet to keep a constant distance and orientation relative to each other by moving the sample or the ionization source and MS inlet. Only a few MSI studies have been performed on large 3D surfaces with sampling MS techniques that applied the concept of aligning the sampling probe with the uneven surface to create 3D distributions and, therefore, are described as 3D MSI. In 2014, Bennett et al. coupled direct analysis in real time (DART) with robotic probe surface sampling via a needle and showed the distribution of different dyes on the polystyrene hemisphere.[4] In 2018, Li et al. used a robotic arm with a needle probe combined with an open port sampling interface (OPSI) and showed the relative abundance of molecules on a plastic coffee cup cover, medicine blister pack, football, and six-well spot plate.[6] In 2021, Ogrinc et al. coupled a robotic arm with the SpiderMass (a water-assisted laser desorption ionization MS technique) for in vivo surgical applications.[17] In this setup, 3D topographical imaging of the surface is combined with molecular surface sampling with the SpiderMass. Topological images and 3D mass distributions were shown ex vivo for a sponge spotted with lipid standards, an apple core with seeds, a skin biopsy, and parts of a whole-body mouse model and in vivo on a human fingertip. Although these prior methods are innovative, they have not been applied to samples that are difficult to sample, like bone and cartilage of joints.

The development of different ambient sampling and ionization mass spectrometry techniques, for in vivo applications, allows for the sampling of sample surfaces without sample preparation independent of the flatness of the sample.[4,16,18] One of these techniques is rapid evaporative ionization mass spectrometry (REIMS),[19] combined with either electrocautery or laser ablation sampling.[20,21] REIMS works by analyzing the smoke containing aerosolized molecules created when electrocautery or laser ablation is applied to biological tissue.[19,21] This smoke is rich in lipids and metabolites and is aspirated to a mass spectrometer via tubing connected to a Venturi pump.[19,21] Schäfer et al. showed that laser desorption ionization MS with a CO2 laser could be applied to biological tissue to generate similar spectra as those created with electrocautery when combined with REIMS.[20] Recently, Genangeli et al. demonstrated the use of a CO2 laser coupled with REIMS for the sampling of bone, cartilage, and bone marrow.[21] Of all the developed ambient techniques, the combination of a CO2 laser with REIMS (LA-REIMS) and the SpiderMass are the only techniques that were demonstrated to be able to sample hard tissue samples, like bone and cartilage.[21,22] LA-REIMS was selected, as this technique is promising for the 3D MSI of samples with large height variations and can be used on bone and cartilage. Laser-assisted REIMS (LA-REIMS) has been used for different tissue classification, metabolic screening, and other applications.[18,21,23−27] LA-REIMS and similar techniques can be used for sampling of 3D surfaces, but the created mass spectra are often not spatially correlated back to the three-dimensional sample location to create distribution images. For LA-REIMS, a high-throughput and automated platform has been developed for the metabolic screening.[18,23−27]

Despite the promising field of applications, LA-REIMS has not yet been applied to the analysis of complex 3D sample surfaces, to the authors’ knowledge. In addition, LA-REIMS currently lacks imaging capabilities, despite the development of automated screening platforms. The coordinates in the three-dimensional space of the sample locations as well as the order of the sampling locations need to be correlated back to the obtained mass spectra per sampling point to enable imaging capabilities for LA-REIMS. Automation of 3D surface sampling with LA-REIMS would allow for (1) increased reproducibility between measurement points, as the distance and the angle between the sampling device and the sample surface is constant; (2) increased precision, as all sampling positions will be defined by a sampling grid with constant distances between measurement points; (3) faster analysis, as the sample or the laser probe is moved to the next measurement position automatically; and (4) the creation of molecular distribution images, as the spatial location of each measurement point can be coupled to the mass spectrum of the sample location. Besides, 3D MSI of the surfaces of, for example, bones and joints, can improve the generation of clinical classification models, as they will be more robust and small changes can be related back to the annotation of the sample surface.

Herein, we describe the development and demonstrate the effectiveness of such an automated device—the “3D MS scanner”—for the sampling of a 3D surface with LA-REIMS. We focus on the comparison between manual analysis of a 3D surface with LA-REIMS and the automated analysis. Furthermore, we show for the first time for LA-REIMS the added value of this automated sampling technique in terms of retaining spatial information on each sampling location, which allows for 3D visualization of the acquired area.

Materials and Methods

Materials and Samples

2-Propanol and leucine-enkephalin were purchased from Honeywell (Seelze, Germany) and Sigma-Aldrich (St. Louis, MO), respectively. A 0.5 mM sodium formate solution was used for calibration of the mass spectrometer.

Apples and cow marrowbone were purchased at the local supermarket for different types of experiments with the 3D MS scanner. The apple and marrowbone were used without any further sample preparation, except for analysis of the bone of the marrowbone for which the remaining soft tissue was removed with a scalpel.

Human femoral heads were collected from five patients that underwent total hip replacement surgeries at Maastricht University Medical Center (MUMC+). The Medical Ethical Committee of the MUMC+ (approval number: 09-2-123) approved the use of “waste” materials from surgeries for scientific research without an additional written informed consent of the patient. The human femoral heads were stored in a −80 °C freezer until analysis and were left to thaw at room temperature before analysis. No further sample preparation was performed.

Surgical CO2 Laser

An AcuPulse Class IV CO2 surgical laser (Lumenis GmbH, Germany) with a wavelength of 10.6 μm was used for all experiments. The laser has a maximum power of 60 W and can be operated in the modalities continuous wave (CW), pulsed wave (PW), and super pulsed wave (SPW). The used laser settings depended on the sample type. An overview of the different laser settings used for the different samples (apple, cow marrowbone, and human femoral head) during the different experiments (comparison between manual and automated sampling, 3D visualization of molecular distributions, and MS/MS experiments) is provided in Table .

Table 1

Overview of Laser Settings for Different Samples during the Different Experimentsa

	comparison				3D visualization				MS/MS
	M	P	D	E	M	P	D	E	M	P	D	E
apple	CW	25	0.1	2.5	CW	25	0.1	2.5	CW	25	0.4	10.0
cow marrowbone	PW	30	0.4	12.0	CW	30	0.1	3.0	CW	30	0.4	12.0
human femoral head	CW	20	0.3	6.0	CW	20	0.1	2.0	CW	20	0.4	8.0

M is the operation mode, P is the laser power (in W), D is the laser pulse duration (in s), and E is the corresponding sampling energy (in J) per measurement point.

Safety Considerations

Appropriate safety measures were put in place, in accordance with local safety regulations when working with a Class IV laser CO2 surgical laser. The laser was placed fully shielded in either the purposely designed and constructed mechanical CO2 laser probe holder or the 3D MS scanner to prevent holding the laser in the hand while it fires. The experimental area was shielded with Class IV laser safety curtains, and all people inside this area wore CO2 laser-grade safety goggles. An anodized aluminum plate was placed underneath the laser beam to prevent scattering of the CO2 laser beam outside of the class II biosafety cabinet. All experiments were performed inside a class II biosafety cabinet that ventilated the created smoke safely. The CO2 laser has a safety interlock to stop the laser on the laser body, which could be pressed in case of emergency.

Manual Setup: Mechanical CO2 Laser Probe Holder

A custom-built, articulated, mechanical arm was used to secure the CO2 laser probe (the end of the laser arm through which the laser beam passes) for safety considerations and to allow for operation of the laser at a fixed position when the laser was fired. This mechanical CO2 laser probe arm has been previously described by Genangeli et al.[21] In short, the arm is made of aluminum alloy 6082, which was glass bead blasted. The moving parts of the arm were kept in place with air-powered breaks, which cause small movements in the position of the arm when releasing them. The laser probe in the holder could be moved by hand to the required position, where it remained while the laser was fired. The laser was fired with a two-button safety box outside of the biosafety cabinet for safety reasons. The sample is placed underneath the laser at a plate at the bottom of the biosafety cabinet.

Automated Setup: 3D MS Scanner Development

The 3D MS scanner was developed to measure the surface of a sample in an automated way using LA-REIMS. The main components of the 3D MS scanner are the different axes, the sample holder, distance sensor, clamps for fixation of the laser probe, and the control box (see Figure ). For the operation of the 3D MS scanner, a sample is placed in a custom-designed sample holder positioned underneath the laser probe for safety considerations. The sample holder is positioned in the middle of the biosafety cabinet. The sample is automatically moved for each measurement point while maintaining a constant distance and angle between the laser probe and sample surface. The aim of this movement is to place the sample underneath the laser probe in such a way that the laser beam is aligned with the surface normal at the measurement point as perfect as achievable within the implemented degrees of freedom. The setup has four degrees of freedom: three translational axes (x, y, and z axes) and one rotational axis (around the x axis). Electric slides (models EGSK and EGSC, FESTO, Esslingen am Neckar, Germany) in combination with stepper motors (model EMMS, FESTO) were used for the motion along the translational axes. A rotary driver (model ERMO, FESTO) was used for the rotational motion with a manufacturer specified repeatability of ±0.05°. A high-performance laser point distance sensor (model BA_OM70, Baumer Holding AG, Frauenfeld, Switzerland) was used for the measurement of the topography of the surface sample. The distance sensor was placed at a fixed distance from the laser probe and has an optimal measurement range of 40 mm. Therefore, the maximum z-height variation in the acquisition areas should not extend this value. The laser probe was held by two magnetic clamps made of the plastic Stratasys-Verowhiteplus that would release the laser probe if a collision between the sample and laser probe occurred to prevent damage to the fragile laser arm. The base frame and sample holder of the 3D MS scanner were made of aluminum and an anodized aluminum plate was attached at the bottom.

Figure 1

Images of the 3D MS scanner. (A) Picture of the 3D MS scanner setup in the biosafety cabinet with the CO2 laser probe inserted. Besides the laser probe, the following elements are indicated: the tubing through which the created smoke is aspirated toward the mass spectrometer (MS), the distance sensor that is used for the topographical measurement, and the sample holder. (B) Design image of the 3D MS scanner with indication of the three translational axes, the rotational axis, and the control box.

The 3D MS scanner was operated using home-built control electronics that were programmed with custom LabVIEW software (version 2020, National Instruments). The control software was running on a Microsoft Surface Pro 6 using the custom-build LabVIEW application MS3D version 2.01.

Analog Laser Trigger Registration with Mass Spectral Data

The position (in three dimensions) of a measurement point must be registered with the correct corresponding mass spectrum for each sampling event for correct processing of the data and the generation of images. Registration of an analogue laser trigger into the mass spectrometer next to the total ion current (TIC) signal is important for this alignment. An eSAT/IN box (D16EST083M, Waters Corp.) was used for the registration of an analog signal during the MS acquisition. An analogue trigger was registered next to the TIC chromatogram in MassLynx (version 4.1, Waters Corp.) during the MS measurement of the sample whenever the surgical CO2 laser was fired. See Figure S1 for an example of an overlay image between the analogue signal and the TIC chromatogram.

Acquisition Process

The acquisition process consists of three steps: (1) homing of the motors, (2) performing a topographical scan, and (3) performing the MS measurement. The three translational motors are homed, and the center of the rotational motor is determined to ensure accuracy and reproducibility at the start of each experiment. The measurement settings and the rectangular acquisition area are defined. Currently, a control software limitation requires all acquisition areas to be rectangles. In the next step, the topographical scan is used to determine the topography of the surface and translate this into motor positions for each measurement point. The topographical scan consists of two phases: (1) a quick scan of the sample surface where a spline is fitted through the points to define the mesh for the measurement points and (2) adjustment of the motor positions at each measurement point based on the mesh and optimization of the distance between the distance sensor and the surface. When the topographical scan is finished, the MS measurement is performed. The control box will send a trigger to the XEVO G2-XS to start the MS acquisition, which also starts the acquisition of the analogue laser trigger signal. The following steps are repeated for each measurement point: (a) the surgical CO2 laser is triggered to fire one laser pulse and an analog trigger is registered at the eSAT/IN box at the same time, (b) the sample is held still for a short waiting period (optimized value: 0.1 s) to allow for the aspiration of the ablated smoke to the mass spectrometer and to prevent disturbance of the smoke by movement, and (c) the sample is repositioned to the next measurement point. A file is generated that contains the x, y, and z coordinates of each of the measurement points at the end of the acquisition. This file is used to assign the 3D coordinates to the mass spectral data of each measurement point in postprocessing.

REIMS-qTOF Instrumentation

The smoke created with the CO2 laser was analyzed with REIMS by aspiration through a flexible tube by a Venturi pump to the source. A benchtop XEVO G2-XS qTOF (Waters Corp., Manchester, UK) equipped with a REIMS source was used for the ionization and detection of the molecules from this smoke. The measurements were acquired in negative ion mode, in sensitivity mode, and with a mass resolution of around 15000 full width at half-maximum (fwhm) at m/z 600. The acquired m/z range for all samples was 100–1500. The instrument was calibrated daily with a 0.5 mM solution of sodium formate. A lock mass solution of leucine-enkephalin (Leu-Enk) in 2-propanol (final concentration 0.075 ng/μL) was continuously infused at a flow rate of 150 μL/min in the REIMS source using a syringe pump. A lock mass solution was constantly infused as it has been shown that infusion of a solvent matrix during the REIMS experiment can significantly improve the ion intensity.[28]

Comparison between Manual and Automated Measurements

Manual and automated measurements of 50 points were acquired to compare the reproducibility for the manual (see the Manual Setup section) and automated (see the Automated Setup section) setups in data acquisition. The manual data acquisition was always performed subsequent to the automated data acquisition, as the manual measurements points were placed in between the automated measurement points to prevent overlap between the sample location of the measurement points. The data acquisition for the manual and automated setups were performed within the same rectangular acquisition area to keep the biological variance in both data sets similar. The comparisons were performed on an apple, the outside surface of a marrowbone (bone tissue), and five human femoral heads. Human femoral heads are molecularly more heterogeneous than apples and bone, and therefore, areas were selected that were homogeneous by eye.

The manual and automated setups were compared based on the overall intensity, the intensity of 10 selected m/z values, and the variance between measurement points. These m/z values were defined per sample to be specific for the sample signal and to cover the measured mass range (Table S1). The overall signal intensity was determined by (1) the maximum TIC peak intensity value and (2) the area of the TIC peak. The maximum TIC peak intensity values corresponding to each measurement point were determined by taking the maxima in the chromatogram trace of the measurement. The TIC peak per measurement point in the chromatogram trace of the measurement was integrated to calculate the area of the TIC peak. The maximum TIC peak intensity value represents the maximum total signal intensity in one scan for each measurement point, while the TIC peak area represents the total signal intensity for the whole measurement point. Therefore, both values are compared between manual and automated measurements. The 10 selected m/z values were compared based on the calculated areas of the extracted ion chromatograms (EICs) for the measurement points. Results are reported as average ± standard deviation (SD) and coefficient of variance (CV,%). The variance between measurement points was compared using principle component analysis (PCA) of the manual and automated setups per sample. The PCA plots can be related to the variance, as the closer the different measurement points (represented by spheres) are together the lower the variance in between them and vice versa.

MassLynx (version 4.1, Waters Corp.) was used for the analysis of the acquired MS data. The coefficients of variances (CV) for the manual and automated data were calculated to determine whether they were significantly different with a coefficient of variance test. The determination threshold for significance was set at p-values <0.05. Abstract Model Builder (AMX) (version 1.0.1581.0, Waters Corp.) was used for the PCA analyses. The mass spectra were binned with a bin size of m/z 0.2 for the mass range of m/z 100–1500. The mass spectra of each measurement point were lock mass corrected, background subtracted and removed, and normalized (TIC normalization) before building the a 10 component PCA model.

3D MSI Measurements

Three experiments were set up to show the use of the 3D MS scanner for the 3D visualization of molecular distributions on different types of samples: (1) 300 measurements points on an apple of which part of the peel was removed, (2) 200 measurement points on the outside of a marrowbone, and (3) 150 measurement points on the outside of a human femoral head. The spatial resolution was set to 2 mm, and the MS acquisition time per measurement point was 5.0 s.

Home-built MATLAB (version 2018b, Mathworks) algorithms were used for data conversion and 3D visualization of the measurements acquired with the 3D MS scanner.

Molecular Identifications

Tentative molecular identifications were performed on the basis of MS/MS and database searches for the m/z values that are displayed in the different figures in the Results. MS/MS fragmentation was performed on a REIMS-qTOF in negative ionization mode with varying collision energies (CEs) between 10 and 40 arbitrary units. Alex123 lipid calculator, METLIN, LIPID MAPS Structure Database (LMSD), and MassLynx (version 4.1, Waters Corp.) were used for the identification of the m/z values. The data sets were lock-mass corrected using mMass (Open Source Mass Spectrometry Tool, version 5.5.0) when the internal lock mass (Leu-Enk) showed a ppm error above 10 ppm. The lock mass corrected m/z values were used for the identifications.

Results

This study consisted of three parts: (1) the characterization of the 3D MS scanner, (2) comparison between the mechanical CO2 laser probe holder controlled by hand and the automated 3D MS scanner, and (3) 3D visualization of the data acquired with the 3D MS scanner.

Characterization of the 3D MS Scanner

The developed 3D MS scanner (see Figure and Figure S2) allows for the mass spectrometer analysis of sample surfaces with large topographical variations in 3D. The whole measurement process of the 3D MS scanner consists of three main phases: homing of the axes for alignment, topographical scan of the sample surface, and MS data acquisition with LA-REIMS. These phases take less than 60 s, 5.1 s per measurement point, and 5.0 s per measurement point, respectively, after optimization. Three hundred measurement points can be acquired within 1 h taking into account the whole process. The sample positions are defined based on the topography of the surface such that there is a constant distance and angle between the sample surface and the laser probe. The spatial resolution is constant in the top view with a distance of 2 mm between measurement points, which is difficult to achieve with manual analysis. The 3D MS scanner allows for the analysis of 3D sample surfaces in an automated and fast way with LA-REIMS. See Supporting Information S1 for further information about the optimization and characterization of the 3D MS scanner.

Comparison between Manual and Automated Measurements

The CVs were calculated as a representation of the reproducibility between measurement points. The reproducibilities of automated acquisitions were compared with the reproducibilities of manual acquisitions for different sample types. Three of the seven CVs were significantly lower for the automated setup using the 3D MS scanner in comparison to the manual setup looking at the TIC peak intensity values (see Table ). In addition, the CVs for the TIC peak intensity for the manual setup for femoral heads 3 and 5 are lower than the CVs of the automated setup, but the CVs are not significantly different. The CVs for TIC peak area values (see Table ) are lower for the automated setup than the manual setup with significant differences in the CVs for five samples. The results of the TIC peak intensities and TIC peak areas show the same trend, with lower CVs for the automated setup. These lower CVs for the TIC peak intensity and TIC peak area show the improved signal reproducibility between measurement points for the automated setup compared to the manual setup. This is further emphasized by the CVs for the selected m/z values, most of the CVs are lower or similar for the automated setup in comparison to the manual setup (see Table S1). The variation between the measurement points for the whole mass range (m/z 100–1500) for both setups can be assessed using PCA plots (see Figure S3). These plots display a smaller or similar variation between the measurement points of the automated setup than between those of the manual setup.

Table 2

Comparison of TIC Peak Intensities between Automated and Manual Measurementsa

	TIC peak intensity
	automated		manual
	avg ± SD	CV (%)	avg ± SD	CV (%)	P-value
apple	5.97E7 ± 1.08E7	18.13	7.99E7 ± 1.47E7	18.46	0.9026
marrowbone	1.27E8 ± 2.66E7	20.90	6.09E7 ± 2.76E7	45.31	<0.0001
femoral head 1	1.25E8 ± 3.24E7	25.88	7.49E7 ± 2.70E7	36.08	0.0347
femoral head 2	2.10E8 ± 3.25E7	15.48	1.02E8 ± 3.79E7	37.23	<0.0001
femoral head 3	1.46E8 ± 3.51E7	23.98	1.19E8 ± 2.14E7	18.03	0.0577
femoral head 4	2.04E8 ± 3.57E7	17.46	1.57E8 ± 3.35E7	21.41	0.1691
femoral head 5	1.53E8 ± 3.21E7	20.99	1.23E8 ± 2.42E7	19.79	0.6916

Comparison of the absolute maximum TIC peak values in the chromatogram per measurement point between automated and manual measurements for the different samples; apple, marrowbone, and five femoral heads. The average maximum TIC peak intensity value is given with the standard deviation as well as the coefficient of variance (CV) as a percentage. In addition, the result of the coefficient of variance test is provided, indicating if the CVs are significantly different (p < 0.05).

Table 3

Comparison of TIC Peak Areas between Automated and Manual Measurementsa

	TIC peak area
	automated		manual
	avg ± SD	CV (%)	avg ± SD	CV (%)	P-value
apple	1.39E6 ± 1.42E5	10.17	1.45E6 ± 2.31E5	15.94	0.0024
marrowbone	3.29E6 ± 3.20E5	9.73	1.35E6 ± 5.60E5	41.48	<0.0001
femoral head 1	3.66E6 ± 7.23E5	19.74	1.52E6 ± 5.08E5	33.49	0.0007
femoral head 2	5.83E6 ± 4.37E5	7.49	2.64E6 ± 1.13E6	42.70	<0.0001
femoral head 3	4.29E6 ± 6.55E5	15.25	2.47E6 ± 4.22E5	17.05	0.4464
femoral head 4	4.82E6 ± 4.57E5	9.48	4.39E6 ± 7.48E5	17.03	0.0001
femoral head 5	4.00E6 ± 4.67E5	11.68	2.74E6 ± 6.09E5	22.24	<0.0001

Comparison of the areas under the TIC peaks in the chromatogram per measurement point between automated and manual measurements for the different samples; apple, marrowbone, and five femoral heads. The areas are determined with integration of the TIC peaks. The average TIC peak area value is given with the standard deviation as well as the coefficient of variance (CV) as a percentage. In addition, the result of the coefficient of variance test is provided, indicating if the CVs are significantly different (p < 0.05).

3D Visualization of Molecular Distributions on Sample Surfaces

An advantage of data acquisition automation with LA-REIMS is the possibility of combining the mass spectra for each measurement point with the coordinates of that point. This allows the study of the molecular distributions of sample surfaces using LA-REIMS.

Comparison of the before (Figures A and 3A) and after (Figures B, 3B, and 4A) images shows regularly spaced patterns of the laser burning spots created during the acquisition. Optical differences in the laser burning spots are assumed to be related to the structural differences of the sample surface for all three samples. It was possible to select m/z values with different molecular distributions that were related to structural differences visible by eye for each measurement. In addition, m/z values could be selected that showed molecular distributions that were related to structural differences not visible by eye. For the apple, m/z 455.4 has a higher intensity in the peel of the apple, m/z 671.5 has a higher intensity in the spots visible in the part without the peel and a bit higher intensity in the peel, and m/z 846.6 has a higher intensity in the parts without the peel (see Figure C). For the outside of a marrowbone, for example, m/z 393.3 seems to be related to the whiter part of the marrowbone with the highest intensities where the white tissue seems to be the thickest, m/z 419.2 seems to be related to the nonwhite parts of the surface, and m/z 717.5 seems to be related to the remaining muscle tissue at the edges of the measurement area (see Figure C). For the human femoral head, different structures can be seen, including a white/soft pink part at the right of Figure A that is the formation of bony tissue due to sclerosis, yellowish tissue in the middle of the image that is a small piece of cartilage, and red tissue toward the stem of the femoral head on the left of the image. For example, m/z 205.9 seems to be more related to the red tissue on the femoral head, m/z 744.5 seems to be related to part of the white, and m/z 1080.9 seems to be related to the more yellow tissue and a bit of the red tissue (see Figure B).

Figure 2

3D visualization of molecular distributions obtained from an apple with the 3D MS scanner. (A) Apple before analysis with LA-REIMS using the 3D MS scanner. Parts of the peel have been removed to show different molecular patterns can be created. (B) Apple after analysis with LA-REIMS using the 3D MS scanner. The total number of measurement points acquired was 300 with a spatial resolution of 2 mm, thus covering an area of 28 × 38 mm (15 × 20 measurement points). (C) 3D visualization of the molecular distributions (absolute intensities) of the imaging experiment on an apple. Three m/z values were selected based on their different distributions: 455.4, 671.5, and 846.6. Tentative identifications of the m/z values can be found in Table S2.

Figure 3

3D visualization of molecular distributions obtained from the outside of a marrowbone with the 3D MS scanner. (A) Outside of a marrowbone before analysis with LA-REIMS using the 3D MS scanner. (B) Outside of the marrowbone after analysis with LA-REIMS using the 3D MS scanner. The total number of measurement points acquired was 200 with a spatial resolution of 2 mm, thus covering an area of 18 × 38 mm (10 × 20 measurement points). (C) 3D visualization of the molecular distributions (absolute intensities) of the imaging experiment on the outside of a marrowbone. Six m/z values were selected based on their different distributions: 128.0, 279.2, 419.3, 585.5, 744.6, and 865.8. Tentative identifications of the m/z values can be found in Table S2.

Figure 4

3D visualization of molecular distributions obtained from part of a human femoral head with the 3D MS scanner. (A) Outside of a human femoral head after analysis with LA-REIMS using the 3D MS scanner. The total number of measurement points acquired was 150 with a spatial resolution of 2 mm, thus covering an area of 18 × 28 mm (10 × 15 measurement points). (C) 3D visualization of the molecular distributions (absolute intensities) of the imaging experiment on the outside of a human femoral head. Nine m/z values were selected based on their different distributions: 205.9, 274.1, 281.2, 642.5, 682.6, 699.5, 744.6, 1008.7, and 1081.0. Tentative identifications of the m/z values can be found in Table S2.

Discussion

Development and Optimization of the 3D MS Scanner

The 3D MS scanner was developed to analyze sample surfaces in 3D using LA-REIMS. Each of the three phases in the acquisition was optimized, and the acquisition rate of the 3D MS scanner is 300 points in approximately 1 h, corresponding to 22.04 cm2 with a spatial resolution of 2 mm. After homing of the axes, the topographical scan and MS acquisition take 5.1 and 5.0 s per measurement point, respectively. This is faster than the surface sampling setups developed by Bennett et al. and Li et al., which were on the order of minutes per measurement point.[4,6] The acquisition time for the 3D MS scanner is slower than the one reported by Ogrinc et al., which was approximately 0.6 s per measurement point with optimized parameters and simultaneous collection of topographical and MS imaging.[17] One of the reasons for the slower acquisition with the 3D MS scanner is the consecutive acquisition of the topographical scan and the MS measurement. Simultaneous collection is not possible with the 3D MS scanner due to the distance between the laser probe and laser point distance sensor. A risk in slower acquisitions could be molecular degradation of the biological samples, but this is expected to be minimal within 1 h. However, molecular degradation should be taken into consideration for the acquisition of larger surfaces, which will have longer acquisition times.

The 3D MS scanner was developed to keep the distance and angle between the laser probe and sample surface consistent. With the implemented four degrees of freedom (three translational axes and one rotational axis) curvatures along the y-axis are fully compensated. Curvatures along the x- and z-axis cannot be fully compensated with these four degrees of freedom, as this would require additional degrees of freedom in the x- and z-axes. An angle of 10° along one of these axes will result in a decrease of only 3% in the laser fluency per measurement point, which is expected to not have a significant effect on the signal intensity in this setup due to the use of a CO2 laser. Bennett et al., Li et al., and Ogrinc et al. used a robotic arm with six degrees of freedom for sampling of the surface.[4,6,17] Six degrees of freedom will allow for better compensation of the surface topography along all axes. Our in-house built setup with four degrees of freedom can still compensate for most topographical height variations, although the possibilities are more limited than for a robotic arm.

Another difference between the 3D MS scanner and previously published setups using robotic arms is the movement of the sample underneath the laser probe for the 3D MS scanner, while in previously published setups the sample lays still. The needle probe connected to the robotic arm needed to move from the sampling surface to DART ion source or the OPSI in the setup of Bennett et al. and Li et al., respectively.[4,6] The SpiderMass probe was connected to the robotic arm together with the transfer line to the mass spectrometer for the robot-assisted SpiderMass setup.[17] The laser probe is moved relative to the sample in the setup of Ogrinc et al., as they aimed to develop a system that could potentially be used in vivo during surgeries in the future. Our main consideration for a setup where the sample moves underneath the laser was safety, as the system is equipped with a Class IV CO2 laser. Another reason for this setup of the 3D MS scanner is the focus on improved understanding of molecular distributions on sample surfaces with large height variations.

Comparison between Manual and Automated Setup

Comparison between the manual and automated setups showed that the acquisition with the automated setup has a higher signal reproducibility between measurement points than with the manual setup. This was shown by the lower CVs for the TIC peak maximum intensities and areas for most of the automated acquisition than for the manual acquisitions (Tables and 3), which is according to expectations. The lower CVs for the automated setup are caused by the higher reproducibility of the tissue ablation by the CO2 laser due to the constant distance between the laser probe and the sample surface with the 3D MS scanner. In addition, the laser probe is placed as perpendicular to the sample surface as possible. Both considerations are more difficult to achieve with manual acquisitions, especially due to small movements of the laser probe holder caused by the air-powered breaks in the manual setup.

The reported CVs for the manual and automated setups are slightly higher than previously reported by Genangeli et al. for the CO2 laser, especially for the manual setup.[21] Direct comparison between the CVs of both studies is complicated by the differences between the studies, which contribute to the CVs. In this study, more heterogeneous samples, especially the femoral heads, more measurement points per experiment, and different sizes of acquisition areas were used in comparison to the work of Genangeli et al. The CVs for the automated and manual setups were based on 50 measurement points, which covers a larger area than the 10 points acquired by Genangeli et al. Therefore, part of the higher CVs might be explained by small differences in the molecular composition and tissue ablation across the area. Another factor contributing to the high CVs can be LA-REIMS, although the whole setup is not changed during acquisitions. LA-REIMS is based on the analysis of the aspirated smoke created by laser ablation. The aspiration of the smoke might be different between measurement points, resulting in variations in the TIC peak intensities and TIC peak areas contributing to higher CVs. In addition, the surface sample will affect the aspiration of the smoke, as the topography of the surface will influence the flow of the smoke. For example, it has been observed that the aspiration of the smoke can be affected for samples with large height variations.

The TIC peak intensities and TIC peak area values were 8.82–58.94% lower for the manual setup than for the automated setup, except for the apple acquisitions. A contributing factor to this might be the different positions of the sample in the biosafety cabinet in both setups. Changes in the aspiration of the smoke were observed based on sample position due to the airflow inside the biosafety cabinet. It is more difficult to keep the same distance between the sample and laser probe as well as the orientation of the laser probe compared to the sample with the manual setup, which can affect the aspiration of the smoke. Decreasing signal intensities with increasing distance between the sample surface and laser probe were shown for the automated setup (see Supporting Information S1). Therefore, a larger distance between the sample surface and laser probe with the manual setup can contribute to the lower signal intensities.

The manual and automated setup both have their advantages. The manual setup is preferred if only a few measurements points, for example, for MS/MS data, need to be acquired, as no homing of the axis and topographical scan needs to be performed, making the manual acquisition faster than the automated one. Besides the higher signal reproducibility, one of the advantages of the automated setup is the larger area of the sample that is available for analysis. For the manual measurements, the sample is placed on a flat surface or could be placed in a clamp. The fixed position of the sample in the manual setup limits the area available for data acquisition, which is not the case for the automated setup. The sample is fixated in the sample holder by tightening the screws for the automated measurements. The sample can be orientated and moved in such a way that the area of interest can be placed underneath the laser probe in the automated setup. The only limitation is that the screws should not be tightened at a position that might be of interest in later acquisitions. Another advantage of the automated acquisition is the automated movement to the next measurement point while maintaining a constant distance between the different measurement points, which prevents any overlap between the laser burning spots of different measurements points. Some overlap between the measurement points for the manual setup occurs sometimes due to the small movements of the holder due to the air-powered breaks. Therefore, the accuracy of the sampling position is higher for the automated setup due to the predefined sampling grit in comparison to the manual setup.

Comparison of 3D Visualization between the 3D MS Scanner and Similar Technologies

3D visualization of molecular distributions in Figures –4 demonstrates that different m/z values could be selected that represented molecular distributions related to visible structural differences but also distributions that were not related to visible structural differences. The spaces between the laser burning spots were equal for the three different analyzed sample types when looking from the top view, but differences between laser burning spots were seen based on structural differences in the surface. Recently, Ogrinc et al. have shown that their robot-assisted SpiderMass setup allows for the visualization of 3D molecular distributions on different samples.[17] The robot-assisted SpiderMass setup has not been applied to human bone and joints, like the human femoral heads used in this study. The SpiderMass is mini-invasive,[17] while the CO2 laser leaves clear burning spots, due to the different equipped laser types. This is a result of differences in interaction between the biological tissue and the laser wavelength as well as different laser energies.

All the acquisitions with the 3D MS scanner were acquired with a spatial resolution of 2 mm. This is larger than the spatial resolution reported by Ogrinc et al., which was 500 μm without oversampling.[17] One of the factors that will limit the spatial resolution is the diameter of the laser focus point. The diameter of the focus point of a CO2 laser is larger than the Nd:YAG laser used in the SpiderMass setup. Therefore, a spatial resolution of 500 μm cannot be reached with a CO2 laser without major overlap between laser burning spots. It is noteworthy that the spatial resolution of the 3D MS scanner could be reduced to the size of the laser focus point, which is approximately 1 mm in this setup.

Future Applications

The 3D visualizations created from the data acquired with the 3D MS scanner allow for observing and correlating changes in molecular intensity to locations on the sample surface and, therefore, structural differences. The 3D correlative images can, for example, be used to study the different molecules present in the different tissue types of a femoral head and the transitions areas between them (see Figure ). This can increase the molecular understanding of the processes underlying the structural changes in, for example, osteoarthritis of the femoral head. Large areas consisting of different tissue types should be acquired of a large cohort of femoral heads to achieve this. Furthermore, this molecular information on a large cohort of samples can be used to improve models build for clinical applications. Another potential application of the 3D molecular visualizations is to determine the optimal sampling spot for tissue classification for in vivo applications of LA-REIMS. On the basis of the molecular distributions, the most informative region for tissue classification can be determined, which can be specifically targeted during in vivo applications. This will reduce the amount of tissue damage created by the CO2 laser as fewer sampling points will be required.

In addition to this specific potential clinical application for femoral heads, the 3D MS scanner has the potential to be used in a wide range of biomedical applications either as is or with small adjustments, for example, to the sample holder. Any type of sample could potentially be placed in the 3D MS scanner if it can be ablated with a CO2 laser, although caution is advised to prevent contamination of the mass spectrometer. Adjustments could be made to include other types of ambient sampling probes that could be triggered in a similar way to also enable analysis of samples that cannot be ablated with a CO2 laser. Two of the biggest advantages of this 3D MSI setup are the lack of sample preparation and the ambient conditions. Sectioning of the sample is commonly needed for 2D MSI techniques, which is not possible for all sample types or can be difficult to achieve. Furthermore, some samples are too large to fit on a microscope slide, which would not be an issue with the 3D MS scanner combined with LA-REIMS. In addition, the ambient conditions can allow for the analysis of samples that cannot withstand the vacuum used in, for example, MALDI-MSI. The 3D MS scanner could have a broad range of 3D MSI applications, for example, in the food industry as well as biomedical and preclinical applications, as shown in Figures –4.

Conclusion

This study describes the development of the 3D MS scanner, which allows for the analysis of large and uneven 3D sample surfaces with LA-REIMS under ambient conditions. The 3D MS scanner automates the positioning of the sample underneath the laser probe and keeps the distance between them constant. This automated acquisition results in increased reproducibility of the obtained molecular information on the sample surface when compared to manual acquisitions. In addition, the data acquired with the automated setup can be used for the visualization of molecular distributions on the sample surface in 3D. This application was shown on an apple, a marrowbone, and a human femoral head, which showed distinct molecular distributions that could be related to structural differences. Therefore, the developed 3D MS scanner can be used to improve molecular understanding based on 3D molecular distributions and might have a potential application in determining optimal in vivo sampling points to reduce biological tissue damage.