Microfluidic Impedance Cytometry for Single‐Cell Particulate Inorganic Carbon:Particulate Organic Carbon Measurements of Calcifying Algae

Abstract Calcifying algae, like coccolithophores, greatly contribute to the oceanic carbon cycle and are therefore of particular interest for ocean carbon models. They play a key role in two processes that are important for the effective CO2 flux: The organic carbon pump (photosynthesis) and the inorganic carbon pump (calcification). The relative contribution of calcification and photosynthesis can be measured in algae by the amount of particulate inorganic carbon (PIC) and particulate organic carbon (POC). A microfluidic impedance cytometer is presented, enabling non‐invasive and high‐throughput assessment of the calcification state of single coccolithophore cells. Gradual modification of the exoskeleton by acidification results in a strong linear fit (R 2 = 0.98) between the average electrical phase and the PIC:POC ratio of the coccolithophore Emiliania huxleyi 920/9. The effect of different CO2 treatments on the PIC:POC ratio, however, is inconclusive, indicating that there is no strong effect observed for this particular strain. Lower PIC:POC ratios in cultures that grew to higher cell densities are found, which are also recorded with the impedance‐based PIC:POC sensor. The development of this new quantification tool for small volumes paves the way for high‐throughput analysis while applying multi‐variable environmental stressors to support projections of the future marine carbon cycle.


Device fabrication
A standard photolithography process was used to fabricate buried 10/135 nm tantalum/platinum electrodes in borosilicate glass wafers (SCHOTT MEMpax, 500 µm thickness). First a lift-off mask was created with photoresist to define the electrode pattern. Then a BHF wet etch was used to make 145 nm trenches, where after the tantalum and platinum layer were sputtered and later removed via liftoff. Lastly, the wafer was diced in separate chips.

Data processing
The differential impedance was measured with a lock-in amplifier (Zurich Instruments HF2LI), recording the real X and imaginary Y part of the signal. The baseline of the real and imaginary signal were removed by subtracting a seconder order polynomial fit and the signal noise was reduced by a moving average filter.
Next, particles/events were detected by the 'findpeak' algorithm of MATLAB. We observe an upward peak followed by a downward peak (or vice versa), owing to the nature of the differential measurement ( Figure 2b). For each particle registration the value of X and Y was determined by the average of the first and the second peak. Then, we find the magnitude (R) and phase ( ) response as follows: The signal response is not merely the result of the passing bead or cell, but it also influenced by the measurement system. E.g., cables will introduce a small time delay compared to the internal reference signal of the lock-in amplifier, which results in a significant phase shift of the baseline signal at high frequency. Therefore, the phase response is normalized with respect to the mean phase response of the reference beads ( according to: . Note, that this normalization protocol differs from our earlier work ( [1]), resulting in a different phase range. The cell diameter was calculated as follows: , where k is calculated using the magnitude response of 5 µm polystyrene beads.
Changes in the measurement system result in a different phase response of the reference beads as can be observed by the beads with the black color in Figure S1. The cells in the 'intermediate' (in red) and 'intermediate -shifted' (in black) run are from an identical treatment, unfortunately even after normalization we still observe a major shift in the phase response ( Figure S1b), indicating that the phase shift of cells with respect to beads is nonlinear. In short, the bead response should be identical within measurements to make a fair comparison. Figure S1: Robustness of the measurements. The measurement in red and black are the same treatment, but a change in the measurement system alters the bead response a) and results in a significant deviation, even after normalization b).
Finally, we discriminate between cells, beads, debris and dead cells ( Figure 2c). Typically particles smaller than 3.5-4.0 µm are classified as debris and ignored, this was verified by optical inspection (e.g. Figure D.2). Cells with a normalized phase response smaller than -0.12 (calibration series) and -0.10 (other measurements) are also removed from the dataset as dead cells ( Figure S3).

Error estimation
We have quantified the measurement error of our system by moving an individual cell and two individual beads 10 to 20 times back and forth over the impedance sensor ( Figure S5a). We can conclude that the biological variation is larger than our measurement error of single cells, based on the comparison of these single particles to a bulk measurement of hundreds of cells and beads ( Figure S5b). We note that the variation in the single cell phase response is larger than of the single beads, possibly owing to different orientations of the possible non-uniformly shaped exoskeleton, which can be asymmetric.

b) Comparison of the single particle measurement error and the bulk measurement of beads and cells. The error bars indicate the standard deviation of the individually measured particles.
CO 2 treatments and cell density additional data Figure S6: Boxplots of the PIC:POC ratio for different cell densities at low and high CO2 (400 and 1000 ppm, respectively). The difference between the low and high CO2 treatment is for none of the cell densities significantly different (t-test; P>0.05). Figure S7: Boxplots of the normalized phase for different cell densities at low and high CO 2 (400 and 1000 ppm, respectively). The difference between the low and high CO 2 treatment is for none of the cell densities significantly different (t-test; P>0.05).