In this study, monodisperse droplets were used to introduce single microbeads (made out of polystyrene) and cells (mouse spleenocytes) into a downward-pointing ICP source. Vertical orientation of the ICP enables quantitative transport of droplets into the plasma independent of the initial droplet size. Analyte elements from the microdroplets were detected via TOFMS and microdroplet-derived signals were registered on the TOFMS time trace by monitoring signal from a droplet tracer element, Cs. For all of the registered microdroplet signals, the signals from 12C+ and other isotopes of interest, including 140Ce, 153Eu, 165Ho, and 175Lu from the REE beads and 31P from the cells, were recorded. Sections of the time-resolved ICP-TOFMS data are provided in Fig. 1. As seen, carbon from the plasma produces a stable background of ~ 75 counts/acquisition independent of any dissolved C present in the droplet; at the sensitivity level of the TOFMS, the empty microdroplets (i.e., droplets without beads or cells) do not contain measurable levels of 12C. Thus, the detection of any 12C+ signal spikes above the pronounced 12C+ background was the result of a bead or cell event. Bead events were recorded and identified based on their spike elements (i.e., Ce, Eu, Ho, and Lu) and cell events were identified based on the coincidence of 31P+ signal with 12C+. In Fig. 1a and b, individual bead events, and, in Fig. 1c and d, individual cell events on the ICP-TOFMS time traces are provided to illustrate how bead and cell events can be separated with multielement ICP-TOFMS analysis. Note that exclusively 12C+ ion signals were considered in this study, as the instrument’s sensitivity was not sufficient to detect 13C+ ion signals quantitatively. Furthermore, the instrument was tuned specifically towards the low mass region, which limited the sensitivity of elements in the high-mass region (above m/z = 80).

Time-resolved signals recorded for monodisperse droplets carrying beads (a and b) and cells (c and d) are shown. Carbon spikes originating from beads and cells were obtained on top of the pronounced background that was found to be stable throughout the acquisition period. Bead events were confirmed according to their spike Ce, Eu, Ho, and Lu and cells according to their P content. b illustrates a close-up of subfigure (a) showing a typical bead event with REEs and Cs as droplet tracer. d illustrates a close-up of subfigure (c) showing a typical cell event with P and Cs as droplet tracer

The sensitivity for 12C+ in ICP-MS is low due to the high first ionization potential of 11.26 eV, which results in an estimated 5% ionization in the ICP [27]. Additionally, inherent mass bias of mass analyzer and ion-transmission optics introduce further losses for measurement of low mass-to-charge (m/z) species [34]. Here, the ion optics upstream of the TOF mass analyzer were optimized for maximum ion transmission of 12C+ which led to a sensitivity loss for heavy elements (i.e., 140Ce+, 151Eu+, 153Eu+, 165Ho+, and 175Lu+) by a factor of five to ten. The sensitivity for 31P+ was comparable with our previously reported scICP-TOFMS results [30].

In Fig. 2, we provide an example histogram of the background-subtracted ion signal distribution for the 12C+ signal distribution for the polystyrene beads introduced via microdroplets. We measured the 12C+ signal instead of 13C+ because the higher sensitivity from 12C+ enabled the detection of all bead events. Bead-derived 12C+ signals were identified based on concurrent detection of 140Ce+, 151Eu+, 153Eu+, 165Ho+, and 175Lu+, which were also in the beads. As the 12C+ background signal was stable throughout a measurement, the net 12C+ signal from each bead (or cell) can be determined by subtracting the mean 12C+ background intensity, which was determined via a Gaussian fit of the background signal histogram. The single-cell dataset was processed in the same manner as the microbead data, with the exception that 31P+ signals were used as the cell tracer.

A fitted and background-corrected 12C+ ion distribution from bead signals is shown

In Fig. 3, we plot the correlation of 12C+ and 31P+ signals from individual spleenocyte cells. As seen, the recorded ion signals for the cell sample yielded broad distributions spanning one and a half orders of magnitude for 12C+ and an order of magnitude for 31P+ signals. This distribution reflects the heterogeneous nature of this spleenocyte sample, which contains different cell types such as B and T cells, monocytes, granulocytes, dendritic cells, and macrophages; all of which show slightly different sizes, and bio-variability [35] 12C+ and 31P+ signals appear to be correlated; as cell carbon content increases, so does the amount of P. This correlation might reflect differences in the cell sizes or, more generally, in the biomass. However, it should be noted that cell volume and biomass are not necessarily linearly correlated with each other [2].

The correlation between 12C+ and 31P+ ion signals (background-corrected) that were obtained for single mouse spleenocytes is shown. An increasing P content for cells with increasing carbon content (and vice versa) was observed

As every carbon spike originating from a single-cell event was accompanied by a 31P+ ion signal, the latter was used to identify and sort out random carbon spikes and signal fluctuations from the distribution of true cell events. The low-count 31P+ background ion signal distribution did not show a normal distribution, but rather followed a compound Poisson distribution [36]. To identify cell signals, we set a manual threshold for 31P+ signals one order in magnitude higher than the obtained average background signal.

An external calibration was carried out using dissolved glucose in microdroplets as a carbon source [27]. Solutions of four different concentrations of glucose were measured by dispensing thousands of droplets into the plasma so that the detected ion signals from individual droplets yielded a normal distribution for each calibration solution. The mean value was obtained by applying a Gaussian fit to each distribution. Figure 4 shows the background-corrected 12C+ ion signals as mean values versus the carbon mass per droplet. A linear regression was carried out and the depicted regression line shows a linear correlation between ion signal and mass up to 50 pg C per droplet with R2 = 0.9988. The limit of detection (LOD) was determined to be 4.83 pg C per droplet.

Calibration curve showing the background-corrected 12C+ ion signal as a function of the carbon mass per sample droplet. A linear regression was carried out yielding R2 = 0.9988 and the regression line with a slope of approximately 7 counts per pg was plotted to visualize the linear relationship for the calibrated analyte mass range. The uncertainty of the carbon mass was derived from the uncertainty of the droplet size by error propagation

As demonstrated by Garcia and co-workers, microdroplets form (semi)-dry, solid particles upon the desolvation of the solvent, which can measure up to several micrometers in diameter [33]. According to the glucose concentrations used in this study, semi-dry glucose particles between 7 µm and 13 µm in diameter would have been expected to form serving as suitable calibration proxies for polymer microbeads and cells. However, here the assumption of forming spheres had to be made.

The regression parameters were used to convert the obtained single-bead and single-cell 12C+ signals into carbon mass per bead and cell, respectively, which is shown in the boxplots in Fig. 5a. Furthermore, the size, i.e., diameter, of the polymer beads investigated was determined and compared with SEM data, which is illustrated in Fig. 5b. A summary of these results is provided in Table 1.

Boxplots showing the carbon mass distribution obtained from single-bead and cell measurements (a). Notably, the carbon mass distribution pattern of the beads is much narrower than the distribution pattern of the cells, reflecting the monodispersity of the beads and the heterogeneity of the spleen cell sample. The bead diameter was derived from the determined carbon mass and the size distribution is shown in (b). SEM was used to confirm the spherical shape and to measure the diameter of randomly sampled beads

In the provided boxplots, the corresponding mean values are highlighted by white dots and the median values by orange horizontal lines while interquartiles are indicated by blue boxes with black rectangles. Red dots are referred to as fliers. The 12C+ ion signals from the beads followed a normal distribution indicated by the shape of the displayed distribution as well as the comparable mean and median values (Fig. 5 and Table 1). The 12C+ ion signals that originated from cells showed a broad distribution, where the median and mean differed by approximately 1.6 pg, which is consistent with the expected heterogenous distribution (naturally given) of the spleenocytes, in terms of cell types and sizes. Notably, the investigated cells and beads yielded a comparable carbon content. Considering the mean values, approximately 15 pg per cell and 13 pg per bead, respectively, were obtained. Although the beads have a much lower volume in comparison to cells, apparently, comparable absolute amounts of carbon can be expected which underlines the general suitability of beads as calibration standards in mass cytometry.

Assuming the beads were made of 100% polystyrene (ρ = 1.05 g cm−3) [37] and exist as spheres, the carbon mass distribution shown in Fig. 5a was converted into a bead size distribution and a mean size of the beads was estimated, which is shown in Fig. 5b. A sample of the same bead suspension batch was investigated by SEM and the diameters of 24 randomly selected beads were determined. A very narrow size distribution was obtained which is shown in Fig. 5b. The bead size determined by ICP-MS is in good agreement with the bead size determined by S(T)EM, as will be discussed below.

The number of cell events (N = 69) acquired via ICP-TOFMS was lower than for the bead analysis (N = 331). A higher number of acquired cell events would have allowed a better statistical description carbon content of the cell population. However, within this proof-of-principle study, we focused on the method development for the analysis of single beads and single cells. We used low cell number concentrations to prevent sampling two individual cells in a single droplet and to reduce the chance of cell agglomeration since cells resuspended in water tend to form agglomerates.

We also investigated the bead size, i.e., diameter, and shape by S(T)EM to validate the bead sizing method via spICP-TOFMS. Results are presented in Fig. 6. The array of several beads surrounded by debris can most likely be attributed to residues (buffer/salt) from the suspension medium. (Fig. 6a) Detailed inspection of a bead and the debris is provided in Fig. 6b where a dendritic growth pattern of, e.g., a salt residue, was observed after drying. In Fig. 6c, rod-shaped debris was observed on the bead surface.

Images of polystyrene beads displaying an array of multiple beads (a), and single beads (b, c) deposited on a Si-wafer which enabled a detailed inspection of the shape and diameter. The beads appeared as spheres surrounded by debris which originated most likely from the suspension medium (buffer/salt) showing a characteristic dendritic growth pattern after drying

A total of 24 beads were investigated by electron microscopy, which is a much lower quantity than the number of beads (331) measured by ICP-MS. The beads showed a spherical shape and appeared to be highly monodisperse with random sampling, which is also reflected by the standard deviation of the SEM data and, additionally, by earlier studies [20, 38, 39]. Thus, given the high uniformity of the beads, the optical analysis of more beads would most likely not have altered the overall statement concluded from the actual dataset, i.e., the agreement of the bead diameters determined by ICP-MS and SEM.

Sizing by quantification of carbon per single bead by ICP-MS with sensitivity calibration via glucose standard microdroplets proved to be a suitable method that allowed a much higher sample throughput than typically observed for S(T)EM. Notably, S(T)EM requires a laborious sample preparation, but provides high-resolution images and thus a remarkable precision. On the other hand, ICP-MS requires almost no sample preparation for polymer beads, but offers significantly lower precision in sizing the particles. This precision is limited by counting statistics and/or droplet-to-droplet signal variations due to axial and radial shifts of the vaporization point of the droplets in the ICP source.