Approaches to increasing analytical throughput of human samples with multi-isotope imaging mass spectrometry

Abstract

Multi-isotope imaging mass spectrometry (MIMS) combines stable isotope tracers with the quantitative imaging of NanoSIMS ion microscopy. With extensive safety precedent, use of stable isotopes in MIMS applications opens the possibility of studying a wide array of biological questions in humans^[1]. Here we describe a series of approaches to increase the effective analytical throughput for detecting rare nuclear labeling events with MIMS. At the level of sample preparation, cells in suspension were either smeared at high density or pelleted cells were embedded and sectioned to reach nuclear depth. Presputtering conditions were optimized for each cell type to ensure the reproducible sampling of nuclei. Adipose tissue posed a different challenge as the large volume of adipocytes results in an obligatorily low density of nuclei in any given plane. Before introducing samples to the NanoSIMS instrument, all nuclei were fluorescently stained, imaged, and their coordinates recorded, allowing automated analysis of fields that contained at least one nucleus and therefore minimizing analysis of dead space. These data emphasize unique challenges posed by human studies, where both ethical and practical issues may limit the administration of stable isotope labels for prolonged periods of time as may be necessary to achieve high labeling frequencies in cells that divide infrequently.

Introduction

Multi-isotope imaging mass spectrometry (MIMS) merges precise measurement of stable isotope tracers with high-resolution ion microscopy^[2]. In contrast to radioisotopes, stable isotopes are innocuous, having an extensive precedent of safe use, including in humans^[1]. Indeed it is the extensive safety precedent with stable isotopes that make human translation one of the most exciting applications of MIMS. While advances in molecular imaging have opened up the possibility of studying metabolic processes in living humans on the tissue level^[3], MIMS can achieve similar objectives on the subcellular level. Here we present strategies to improve analytical throughput of human tissue where the experimental objectives require sampling a large number of nuclei. We describe optimization of the analytical strategy at the level of both sample preparation and instrumental analysis.

Methods

NanoSIMS

Analyses were performed using the NanoSIMS prototype or the NanoSIMS 50L (Cameca). Both instruments use a Cs⁺ ion source. Samples were polarized at -8 kV and the Cs⁺ ions hit the sample with a total energy of 16 kV. Analyses were performed on square fields and the dimensions reported in the figure legends. Ratio images are displayed using a hue saturation intensity transformation with the lower bound of the scale set at natural abundance (N¹⁵/N¹⁴=0.0037). For ease of viewing, all nitrogen ratio scales are multiplied by a factor of 10⁴, such that 0.0037 is reported as 37.

Human studies

Studies were reviewed and approved by the Partners Institutional Review Board. ¹⁵N-thymidine (Cambridge Isotope Laboratories, Inc) was administered in 0.9% NaCl with a 30 mg bolus loading dose, followed by a rate of 15 mg/hr.

Mouse studies

C57B16 male mice (8-10 wks-old) were administered a single dose of ¹⁵N-thymidine 500 μg and sacrificed 4hrs later.

Cells

For analysis of white blood cells (WBC), the buffy coat was collected and treated with red blood cell lysis buffer (Invitrogen). The resultant WBC suspension was washed with PBS, and fixed with 4% paraformaldehyde (PFA). A small aliquot of WBCs in a volume of 10 μl was applied to the surface of a silicon chip and the pipet tip was used to gently smear the cell suspension over the surface. Oral epithelial cells were obtained with a cytobrush, smeared on a silicon chip, and fixed with 4% PFA. All smeared cells were alcohol dehydrated and air-dried. The desired cell density was verified using differential interference contrast microscopy.

Tissue sections

After fixation (4% PFA), tissues were embedded in LR white (intestine, fat) or Epon (pelleted white blood cells), sectioned to 0.5 μm, and mounted on silicon chips. When indicated, DAPI (Invitrogen) was applied to sections for 15 min, followed by 3 washes in PBS. Photographs of DAPI-stained nuclei were captured and the coordinates recorded.

Results

Scanning for rare labeled cells in smeared cell suspensions

A major focus of prior MIMS studies has been on the use of stable isotope-tagged thymidine as a marker of cell division^[4,5], but for many cell type division is infrequent and therefore labeling frequency is low. Analysis of leukocytes obtained from a human volunteer who received ¹⁵N-thymidine for 48 hours by intravenous infusion presented such a challenge^[4]. To increase the number of cells per analytical field, leukocytes were isolated from the peripheral blood and smeared at high density on silicon chips. Preliminary analyses suggested that presputtering a field size of 80 μm at 1.5 nA for 10 min reproducibly reached the level of the nuclei in the majority of smeared leukocytes. Therefore, to maximize throughput, we performed automated chain analysis using these analytical conditions, which enabled overnight analysis. Shown (Figure 1) is a representative mosaic image of several hundred leukocytes, yielding a single ¹⁵N-thymidine labeled cell and which was confirmed with higher-resolution imaging.

Figure 1. Rare ¹⁵N-thymidine-labeled human white blood cells.

(a) Mosaic HSI image of smeared peripheral white blood cells shows a large field with one ¹⁵N-thymidine labeled cell (arrow). Presputtering (1.5 nA) for 10 min/field reproducibly reached the level of the nucleus in the majority of white blood cells. Field = 80 × 80 μm, 5 planes. Pixels = 128 × 128, 4.0 ms/pixel. (b) Additional analysis at higher resolution confirms ¹⁵N-labeling (top cell) with an adjacent unlabeled cell (bottom cell). Field = 12 μm, 4 planes. Pixels = 256 × 256, 2.0 ms/pixel.

Oral epithelial cells posed a more significant analytical challenge due to their large volume and relatively small nucleus, which obliged us to use a large field size and therefore lower current density during presputtering. Although the nuclear contours were evident in CN images (Figure 2) after 10 min of presputtering (1.5 nA, field = 120 × 120 μm), the absence of a strong P signal at that depth indicated the absence of nuclear material in the analytical plane. Indeed, a significant P-signal was not observed until approximately 20 min total presputtering at 1.5 nA. These analytical conditions were used in an automated chain analysis, which successfully reached the nucleus in 9/9 analyzed cells.

Figure 2. Optimizing sputtering conditions to sample the nuclei of smeared oral epithelial cells.

Oral epithelial cells swabbed from the buccal mucosa of a healthy human volunteer. Left: Top row shows epithelial cell cluster. Nuclear contours are evident (arrows) in the CN^- image, although the nucleus is not sampled in this plane (negative in P^-). Field = 120 × 120 μm, 3 planes. Pixels = 256 × 256, 2.0 ms/pixel. Middle/bottom rows focus on a single cell, sputtered to reach the nucleus. Chromatin, visible in the P^- image, appears dark in the S^- image. Field = 70 × 70 μm, 4 planes. Pixels = 256 × 256, 2.0 ms/pixel. Right: These presputtering conditions informed automated chain analysis. Presputtering for 25 min (1.5 nA) reproducibly reached the nucleus (arrow) in 9/9 analyzed cells. Field = 80 × 80 μm, 1 plane. Pixels = 256 × 256, 2.0 ms/pixel. Higher resolution analysis (bottom) confirms chromatin (p^high/S^low). Field = 20 × 20 μm, 2 planes. Pixels = 256 × 256, 2.0 ms/pixel.

Analysis of embedded and sectioned cell pellets to minimize time lost to presputtering

To directly analyze nuclei, we analyzed sections of pelleted and embedded cells (Figure 3). During sectioning, traditional histologic stains, such as hematoxylin and eosin or toluene blue stains can be used to confirm that the level of the nuclei has been reached before mounting sections on silicon chips. Analysis of sectioned human white blood cells show a high density of nuclei as confirmed by the high phosphorus signal emanating from chromatin.

Figure 3. NanoSIMS analysis of sections of pelleted cells obviates the need for prolonged presputtering.

Peripheral white blood cells obtained from a human volunteer after intravenous ¹⁵N-thymidine. The cell pellet was embedded in Epon and 0.5 μm sections mounted on silicon chips. Sections were imaged in automated chain analysis mode with tile dimensions of 30 × 30 μm. One representative tile is shown. This strategy resulted in sampling at the level of the nucleus in >50% of imaged cells. Granules contained in granulocytes are seen brightly in the S^- image. Two ¹⁵N-thymidine labeled cells (arrows) are seen in the ¹²C¹⁵N^-/¹²C¹⁴N^- HSI image. Pixels = 256 × 256, 2.5 ms/pixel.

Increasing the frequency of nuclear sampling in histological tissue sections

Analysis of sections of solid tissues with a low concentration of nuclei per volume, such as adipose tissue requires a different approach (Figure 4a). Building on prior work combining offline fluorescent imaging data with MIMS analyses^[6], here we describe an approach to identify nuclei before the samples are introduced to the NanoSIMS instrument, allowing for targeted nuclear measurements (Figure 4b). Sections were DAPI-stained, allowing the recording of nuclear coordinates before the samples were introduced to the instrument. The coordinates were then used to perform automated chain analysis, where despite narrow field dimensions, each field contained at least one nucleus.

Figure 4. Targeting regions of interest with fluorescent nuclear staining prior to NanoSIMS analysis.

(a) Human subcutaneous fat. Adipocytes are among the largest human cells, however, their volume is primarily due to a single large lipid droplet (appearing black in ¹²C¹⁴N^- image). For any given plane the nuclear density is low (white arrows show nuclei in ³¹P^- image). Mosaic images constructed from 50 × 50 μm tiles, 20 planes per tile. (b) To minimize analysis of fields containing no nuclei and to facilitate higher resolution imaging of nuclei, the coordinates of all nuclei (DAPI) were recorded prior to NanoSIMS analysis. ¹²C¹⁴N^- image shows stromal-vascular cell (small arrow) and an adipocyte nucleus (hatched arrow). (c) Though DAPI contains nitrogen, it did not result in a visible dilution of the ¹⁵N-signal. Mouse small intestinal crypt harvested after 500 μg I.P. ¹⁵N-thymidine. (d) Bar graph showing a trend (t-test, p=0.09) toward dilution (∼5%) of the ¹⁵N-signal after DAPI (control = adjacent sections, no DAPI).

We considered the possibility that DAPI, which contains nitrogen (C₁₆H₁₅N₅), would dilute the ¹⁵N-signal. We therefore measured the effect of DAPI staining on the ¹⁵N-signal in serial sections of ¹⁵N -thymidine labeled mouse small intestine (Figure 4c). Although there was a trend towards signal dilution in DAPI treated sections compared to control unstained adjacent sections (Figure 4d), the relative reduction in signal was ∼5%, which is unlikely to bias the results with the high level of labeling achieved with exogenous administration of stable isotope labels.

Discussion

Here we describe several strategies to increase analytical yield of human tissue samples analyzed with MIMS, where the experimental endpoint required sampling of nuclei. The experience detailed here underscores the larger theme that each new experimental application of MTMS requires a “custom” approach to both sample preparation and instrumental analysis to achieve the analytical goals. While the question of analytical throughput is of broad relevance to the NanoSIMS community given the demand for instrument time, human studies pose unique challenges worth considering. The first is that the expected inter-subject variance in a given outcome is generally higher compared to experiments performed in genetically homogenous model organisms housed in carefully controlled environments; therefore, human studies will generally require more subjects to achieve statistical power. In addition, for studies involving stable isotope labeling, cost and other practical issues often limit the duration of label administration. For example, while we have administered ¹⁵N-thymidine to mice continuously for months^[4], such a strategy would be prohibitively costly and impractical in humans. This underscores the larger importance of developing strategies to maximize analytical throughput in order to achieve the ultimate goal of widespread human translation of MIMS.