1. Introduction

The total refractive index of a medium is represented by a complex number n + ik, where the imaginary part of the refractive index, k is determined by the absorption of the medium and the real part, n is defined as the ratio between the phase velocity of light in a vacuum and in the medium. Both n and k affect light scattering, but n plays a primary role [1]. Furthermore, n of a medium or particles is closely related to their density, which in turn is determined by composition. Therefore, the real part (n) of the complex refractive index of particles, which will be referred hereafter as refractive index (RI), provides information on their composition. Bio-optical models using remote sensing observations have been instrumental in advancing our understanding of the significance of marine phytoplankton to the global carbon cycle [2]. However, bio-optical models of the scattering properties of phytoplankton and other particles are still limited by a lack of information regarding their RI [3]. Accurate estimates of RI are, therefore, fundamental to quantify the interaction of particles with light and improve their detection from remote sensing observations.

In 1996, Aas [4] reviewed the RIs estimated for different mineral particles, bulk particle populations, and phytoplankton species in the marine environment using a variety methods, including immersion [5], phase contrast [6], flow cytometry [7–9], angular scattering of the volume scattering function [10–14], angular scattering of Mueller matrix [15], and attenuation and absorption efficiencies [16–19]. He found that the RI values vary with the method used and even within methods dependent on their assumptions. For example, Bryant et al. [20] measured the RI value of spinach chloroplast using two methods and obtained values at 589 nm of ∼1.43 with the immersion method and ∼1.37 with the attenuation method. Kullenberg [13], Gordon and Brown [10], Brown and Gordon [11], and Zaneveld et al. [14] used the same volume scattering function that Kullenberg [21] measured in Sargasso Sea to estimate RI for the same bulk particle population, but obtained vastly different RI values by assuming different size distributions. Additional challenges for methods based on light scattering, absorption and/or attenuation is that they often assume particles are homogeneous spheres to facilitate Mie computation. Natural marine particles, however, are rarely spherical or homogenous. Recognizing these methodological uncertainties, Aas estimated the bulk average RI of phytoplankton as the volume-weighted average of the RI values of their metabolite composition, whose organic components included protein, carbohydrate, lipid, and pigment, and inorganic components included silica and calcite. He first estimated the average RIs for the dry mass composed of these cellular components and then estimated the bulk RI of phytoplankton cells as a mixture of dry mass and cellular water content. He found that the bulk RI values vary across phytoplankton taxa and with the water content of phytoplankton cells, with the latter being a dominant factor. For example, the biggest difference in RI of phytoplankton with a 60% water content was found between coccolithophores (1.424) and brown algae (1.414), and with an 80% water content this difference shrank to 1.381 versus 1.376. Although this pioneering work by Aas (1996) is widely cited (245+), there have been few attempts to verify his estimates. Furthermore, although water content appears to be an important factor determining the bulk RI of phytoplankton, little is known about how it varies with cell size.

It has long been recognized that cellular structure can significantly affect light scattering, especially backscattering, by phytoplankton cells [22,23]. Indeed, recent studies have demonstrated that using coated sphere models, with shell representing the cell membrane and the core the internal organelles of phytoplankton, could simulate the optical properties of phytoplankton cells agreeing better with observations than the homogeneous spherical model [24–31]. In these studies, the RI values assumed or derived typically ranged from 1.44 to 1.60 for the shell and 1.36 to 1.42 for the core. The higher RI values for the shell compared to the core reflects the fact that the shell is typically made of silicate, calcite, cellulose, and/or protein for which RI values are much greater than water/cytosol, which is assumed to be the dominant component of the core. The coated sphere model also requires knowledge of shell thickness. Some studies assumed or derived shell thickness varying from 0.05 to 0.4 µm [24,26,29], approximately overlapping the ranges of thickness of frustules [32] and coccoliths [33]. Other studies have also used chloroplast to represent the shell, whose volume was assumed to constitute 15% - 20% of the total cell volume [25,31]. The use of coated sphere models to simulate the optical properties of phytoplankton cells still requires further validation [34], but their burgeoning success prompts the need to improve our understanding of the RI of phytoplankton cellular structures and their volumetric distributions.

The 3D Cell Explorer (Nanolive SA, Switzerland) is a holographic and tomographic laser microscope that generates a 3D image of a sample with contrast dictated by the differences in the refractive index [35]. It permits the internal structures of a particle to be visually examined at a resolution below the optical diffraction limit (< ∼0.2 µm). The objective of this study was to apply this technological advancement to measure RIs and volumes of various cellular structures of live plankton. The significance of this work is two-fold (1) it provides an independent method for measuring RI to compare to previous work, and (2) provides measurements of the RI of the essential constituents of plankton and their proportional volumes, which pave the way for improving the simulation of the optical properties of these particles and their retrieval from optical observations.

2. Material and methods

2.1 3D Cell Explorer

3D Cell Explorer applies holographic, rotational scanning [36], and tomographic techniques to provide a three-dimensional (3-D) image of a cell constructed from measured refractive indices of different cellular components [35]. The instrument shines a laser at 520 nm, which is subsequently split into two beams: a reference beam and a sample beam (Fig. 1). While the reference beam travels unimpeded, the sample beam penetrates the cell at an angle (45° for our instrument) and undergoes a phase change due to its interaction with the cell. A second beam splitter recombines the sample and reference beams, forming an off-axis interference pattern or hologram at the image plane. The phase contrast signal used to generate the hologram depends on the distance that the light has traveled and the difference of RIs along the path between the cellular constituents and the background [37]. Hence the hologram contains integrated information on the location of a pixel and its refractive index, both of which are retrieved later through a deconvolution algorithm. As the sample beam rotates (∼1.5° per angular rotation), a charge-couple device (CCD) camera records the hologram at 13.3 frames per second (via a software called STEVE), providing a total of 240 holograms in 18 s for 360°. After one 360-degree view, the holograms at various angles are deconvoluted iteratively by a tomographic algorithm [38] (also via STEVE program) to create a 3D cell image with each pixel value representing the refractive index relative to the medium of the cellular content at that location. Additionally, the instrument utilizes a specially designed “high-numerical-aperture” to collect light using many small apertures rather than one aperture like a typical microscope [39], which increases the resolution up to two-fold. The instrument has a total field of view of 80 µm × 80 µm × 30 µm, a lateral resolution 0.2 µm, an axial resolution 0.4 µm, and an RI resolution 0.001. In the STEVE software, different RI values can be color-coded for a detailed examination of various parts of a cell in 3D.

 

Fig. 1. Diagram illustrating the basic working principle of the 3D Cell Explorer. A laser beam at 520 nm is split into two beams: reference and sample. The sample beam, after penetrating the cell at an angle, is recombined with the reference beam, producing a holographic image. The sample beam is rotated 360 degrees, with resulting holograms at each angle are deconvoluted by tomographic algorithms to produce a 3D image of the cell.

Download Full Size | PDF

For live samples, the working lower size limit of particles that can be clearly imaged was found to be approximately 2–4 µm. Smaller cells or particles, such as bacteria, are moving relatively rapidly under Brownian motion. While they can be seen clearly in 3D Cell Explorer, clear 3D images necessary for analysis were difficult to obtain. The upper size limit (<20 µm) was mainly determined by particle/cell thickness and the limited range of the microscope adjustments to the field of view (i.e., vertical range of microscope stage). Although the technical specifications of the microscope state an absolute RI resolution of 0.001, an RI difference of ≥0.005 was necessary to accurately distinguish among different cellular structures.

2.2 Calibration

RI values directly derived from the 3D Cell Explorer are within a range of approximately ±0.1 of the RI of the background medium. In the present study, filtered seawater was used as the medium and assigned an RI equal to 1.339. Therefore, directly derived RIs are in the range of 1.239 to 1.439. This does not mean that RI values outside this range cannot be resolved, but rather actual RI values are projected into this narrow range dictated by the background medium. To estimate actual RI values, we used several standards to develop a calibration curve for correcting measured RI values.

Three standards were obtained from suitable sources: lab grade silica dioxide for silica, a coral fragment for the calcium carbonate (calcite), and a sonicated leaf for the cellulose. Each of these samples was ground into fine particles and suspended in filtered seawater (RI = 1.339). Figure 2 shows the resulting calibration curve (R² > 0.99) from the RI values measured for the three standards in comparison to their reference values compiled in Aas (1996). Specifically, we used the RI value 1.457 measured for Skeletonema [40] as reference value for silica, RI value 1.601 [41] for calcite, and RI value 1.55 [42] for cellulose. The RI value for seawater (serving as the medium) was also measured and included. The associated regression equation was used to correct measured RI values. Hereafter only corrected values are reported. These corrected values represent the absolute RI values or RI values relative to vacuum.

 

Fig. 2. The RI values measured with the 3D Cell Explorer for three standards, SiO₂, CaCO₃, and cellulose are compared with their reference values reported in Aas [4]. The RI value for the filtered seawater is used as an additional data point.

Download Full Size | PDF

2.3 Field samples

Water samples were collected in May 2021 within an anticyclonic eddy centered at 50°N, –15°W in the North Atlantic during the NASA EXPORTS cruise onboard the RRS Discovery. Samples were collected from Niskin bottles attached to the CTD rosette. After collection, samples of approximately 200–500 mL were concentrated to 10 mL over 10–20 µm mesh, and examined in 50 µL aliquots using the 3D Cell Explorer.

Water samples were also collected during multiple cruises which took place in the coastal waters of Mississippi onboard RV Jim Franks in February of 2022. During these cruises, 20 L of surface water was collected at various stations, stored in the dark at ambient temperature and transported to shore (within a few hours of collection). In the lab, 500 mL–1 L of each sample was removed and stored overnight at 4°C. After overnight incubation, 200 mL aliquots were removed and concentrated to 4 mL by vacuum filtration using 2 µm pore-size Nucleopore Tract-etched Membrane (Whatman Lab) at minus 0.4–0.2 bar of pressure. Samples were filtered using a single or multiple filters (in 50 mL amounts) to prevent significant increases in filtration pressure (i.e., filter clogging). Cells were then removed from filters by rinsing with filtered seawater into a 15 mL centrifuge tube. For analysis, 25 µL of each sample was transferred onto a Thermo Scientific (Cat. No. 3011) microscope slide and topped with cover slip (Fisher Scientific 12544A).

For data processing, 3D images were loaded into the STEVE program and the background medium, external membrane, chloroplast (if present), and internal fluid of particles were measured, and the volume of each constituent estimated as the product of the number of pixels and the volume of a pixel.

2.4 Emiliania huxleyi cultures and experiments

Coccolithophores were not observed in samples from either of the two field experiments. Due to their ecological and global importance to the carbon cycle [43] and their significance to remote sensing reflectance [44], cultures of E. huxleyi were obtained from the Bigelow culture collection. E. huxleyi monocultures also provided an opportunity to chemically manipulate live cells to remove their coccoliths, thereby directly demonstrating the change in RI as a result of a change in membrane composition from the CaCO₃ coccolith to cellulose membrane that lies beneath.

Monocultures of E. huxleyi coccolithophores (CCMP 371) were grown inside 25 mL culture flasks (Corning Incorporated, 353107) in L1-Si/25 media (Guillard & Ryther 1962) within an environmental chamber (Sanyo Scientific, MLR-351/351 H) at 20°C and under a 12:12 light cycle of 50 µMol m⁻² s⁻¹ (measured with a Quantum Meter from Apogee Instruments). Growth curves were generated based on observations of abundance measured using a BD FASCelestra BVR flow cytometer with blue 488 nm argon laser and the 600LP and 695/40 emissions filters. Growth measurements were made once a day at 9 am. E. huxleyi populations were distinguished using bivariate plots of side scatter and 695/40. Cells were allowed to acclimate to growth conditions for 8 weeks before experimentally manipulated.

For experiments, cells were inoculated into duplicate 25 mL culture flasks containing fresh media at an average starting concentration of 6.7 × 10⁴mL⁻¹ and growth tracked using flow cytometry (FCM). Stationary phase was reached at 3 days of growth and a concentration of ∼5.5 × 10⁵ cells mL⁻¹. On day 5, cultures were placed in darkness for 36-48 hours. This darkness phase depletes phytoplankton of excess energy and prevents them from rebuilding their coccoliths following acidification. From this step onwards, everything was performed under very low light levels to limit growth. Cultures were then split into two parallel treatments, one control series (untreated) and one series treated to remove the coccolithophores. To decalcify the coccolithophores, duplicate cultures were treated with 2.5 mL L⁻¹ of 1 M hydrochloric acid (HCl) [45,46] and gently mixed for 1 minute, resulting in the pH of cultures being reduced to 4.8. Cultures were then neutralized with 1 M sodium hydroxide (NaOH) to their pre-acidification pH (7.8). Immediately following neutralization, each sample was measured via FCM to determine cell loss. The acidification/neutralization treatment resulted in negligible losses in cell counts. For RI measurements, a total of 150 different particles were analyzed from each of the two treatments, acidified and untreated.

3. Results

3.1 Field samples

Phytoplankton were the dominant organisms observed using the 3D Cell Explorer. Diatoms were the most abundant phytoplankton observed in field samples (N = 728), specifically members of the genus Skeletonema (N = 297) and Chaetoceros (N = 208) were the most common, with Pseudo-Nizschia spp., Bacteriastrum spp., Asterionellopsis spp., Thalassionema spp., and Eucampia sp. also being present. In the North Atlantic, Chaetoceros spp. appeared to be dominant, whereas in the Mississippi coast waters, Skeletonema spp. were the most abundant. Dinoflagellates (N = 23) and ciliates (N = 13) were also measured and distinguished by their size, shape, and distinct morphological characteristics. Some species were distinct among the different sampling locations: Cylindrotheca sp., Guinardia sp. and ciliates were only observed in the North Atlantic while Asterionellopsis spp. and dinoflagellates were found only in the Mississippi coastal waters.

Figure 3 displays an example of color-coded images of several phytoplankton species and one ciliate. Each color represents areas of measured RI values, within a range of mean ±0.005. Figure 3(A) and (B) correspond to one cell and a chain of cells, respectively, of Chaetoceros spp. in girdle view, Fig. 3(C) – (G) correspond to the girdle view of Skeletonema spp., Guinardia sp., Asterionellopsis spp., Eucampia sp., and Thalassionema spp., respectively. The blue color in Fig. 3(A) – (G) corresponds to the silica frustule with a mean RI value of 1.453, 1.467, 1.441, 1.457, 1.467, 1.437, and 1.427, respectively; and the green color corresponds to the chloroplasts with a mean RI of 1.483, 1.520, 1.493, 1.479, 1.517, 1.491, and 1.493, respectively. The remaining two images in Fig. 3 illustrate two non-diatom species that were observed. Figure 3(H) is a Prorocentrum sp. dinoflagellate with the blue color corresponding to its cellulose membrane with a mean RI of 1.560 and green portion corresponding to chloroplast with a mean RI of 1.511. Figure 3(I) shows a marine ciliate observed in a North Atlantic sample. In this image, the blue color corresponds to the phospholipid membrane with a mean RI of 1.379, the green color corresponds to its cytoplasm with a mean RI of 1.510. In all the images shown, cytosol is not color-coded, as the color would spill over to the background medium, suggesting that the RI of cytosol is very close to that of filtered seawater (RI = 1.339).

 

Fig. 3. RIs of plankton collected in the field: (A) Chaetoceros spp. in girdle view, (B) Chain of Chaetoceros spp. in broad girdle view, (C) Chain of Skeletonema spp. in broad girdle view, (D) Chain of Guinardia sp. in broad girdle view, (E) Chain of Asterionellopsis spp. in girdle view, (F) Eucampia sp. in girdle view, (G) Thalassionema spp. (H) dinoflagellate Prorocentrum sp., (I) marine ciliate. The different colors correspond to different cellular structures: (A-F) blue: silica frustule, green: chloroplast; (G) blue: cellulose membrane, green: chloroplast; (H) blue: phospholipid membrane, green: internal structure; (I) blue: outer boundary, green: internal composition distinct from the surrounding seawater. Each color represents RI values within a range of mean ±0.005.

Download Full Size | PDF

3.2 RI of Emiliania huxleyi coccolithophores

Flow cytometric analysis demonstrated the effectiveness of the acidification treatment to remove the CaCO₃ coccolith, as populations displayed markedly different side scatter (Fig. 4). Figure 5 shows RI images of E. huxleyi cells in untreated and acidified cultures. The calcium carbonate coccolith was found to have a mean RI = 1.582 (Fig. 5(A), blue color), which was significantly higher than the cellulose membrane (RI = 1.546) exposed after acidification (Fig. 5(B), yellow color). The RI of chloroplast varied little between treated (1.509) and untreated (1.516) cells (Fig. 5(A) and (B); green color). Free floating coccoliths were observed in samples from the untreated cultures, but were not present in the acidified sample, further confirming the effectiveness of the acidification treatment. The RIs of the cellulose membrane of E. huxleyi differ from the RIs of coccoliths beyond the detection resolution, and therefore can technically be resolved on an untreated cell. However, the membrane was difficult to detect when coccoliths were present (Fig. 5(A)). Among 150 images of untreated E. huxleyi cells examined, the cellulose membrane underlying the coccolith was only observed in 3 images but was easily observed following acidification (Fig. 5(B)). The RI values of cellulose membrane of the untreated cells were comparable to those of the acidified cells. Acidification also reduced the diameter of measured E. huxleyi cells from 6.70 ± 0.166 µm to 5.83 ± 0.24 µm.

 

Fig. 4. Flow cytograph displaying a mixed community of treated (acidified) and untreated (unacidified) E. huxleyi populations. The decrease in side scatter is due to the dissolution of the calcite coccoliths after acidification.

Download Full Size | PDF

 

Fig. 5. 3D Cell Explorer images of E. huxleyi samples. A. An untreated sample, with blue corresponding to calcium carbonate and green representing the chloroplast. B. An acidified sample, with yellow corresponding to the cellulose framework that lay beneath the external coccoliths and green corresponding to the chloroplast.

Download Full Size | PDF

3.3 RI values for different cellular structures

The distribution of RI values measured for various cellular components of five plankton groups are shown in Fig. 6 and summarized in Table 1. Figure 6(A) shows the RI values measured for the outer membrane of these five groups: SiO₂ frustule of diatoms, cellulose membrane of dinoflagellates (Prorocentrum sp.), CaCO₃ coccoliths of E. huxleyi, cellulose membrane of the acidified E. huxleyi, and phospholipid membrane of ciliates. The RI values measured for the diatom frustules exhibited the largest variation, likely reflecting the wide range of diatom species observed in both North Atlantic and coastal Mississippi samples. Dinoflagellates observed in Mississippi coastal samples also showed a lot of variability, which may be due to the low sample size of dinoflagellate measurements (N = 23). The RI values of their cellulose membrane were not significantly different from those of acidified E. huxleyi cells, and their respective mean values are nearly identical (Table 1). The RI values of all the membranes, except those of dinoflagellate and acidified E. huxleyi, were statistically different (p value < 0.05).

 

Fig. 6. The distribution of RI values measured for six representative organisms/particles grouped in (A) for external membranes, (B) for Chloroplast, and (C) for non-chloroplast internal fluid.

Download Full Size | PDF

Table 1. Refractive Index values (mean ± standard deviation) measured for various cellular structures. The reference values were obtained from Aas (1996).

View Table | View all tables in this article

The measured RI for chloroplast displayed much higher within-group variability than the membrane (Fig. 6(B)). This may suggest greater intercellular variability in RI for chloroplast than membranes. However, the variability across taxa and mean RI values between groups was similar, i.e., chloroplast RI values were not statistically different among the four phytoplankton taxa: diatom, dinoflagellate, E. huxleyi, and acidified E. huxleyi (Table 1), suggesting that the composition of the chloroplast membrane does not vary significantly among the different phytoplankton species. The major internal structure of ciliates had RIs close to those of chloroplast (Fig. 6(B)), which suggests that this might represent the bulk RI of the cytoplasm. Cytosol RI values varied little among different species (Fig. 6(C) and Table 1) and were not statistically different from the RI of the seawater medium (1.339).

3.4 Fractional volume of cellular structures and bulk RIs

We estimated the volume of each cellular structure by counting the number of pixels identified for each structure and multiplying by the pixel volume. The volume of cytosol was difficult to estimate due to their RI being very similar to the RI of the background medium and resulting in background pixels mistakenly identified as cytosol. To overcome this, the total volume of cells was first estimated. Specifically, we estimated the cross-section area of a cell using a regular shape that best approximated its perimeter. The regular shapes we used include ellipse, rectangle, and triangle. From the estimated cross-section area, the equivalent circle diameter was estimated and used to calculate the volume of entire cell assuming the cell is spherical. The volume of cytosol was then estimated as the difference between the total volume and the volumes of other non-cytosol cellular structures. Figure 7 shows the resulting volume fractions estimated for outer membrane, chloroplast, and cytosol of different phytoplankton species that were measured.

 

Fig. 7. Volume fractions estimated as a function of equivalent sphere diameter (ESD) for (A) membrane, (B) chloroplast, and (C) cytosol of different phytoplankton species that we have measured. The black lines are the regression. The mean and standard error of regression are 3.72 ± 0.24, 1.18 ± 0.59 in (A) and 1.03 ± 0.06, –2.28 ± 0.24 in (B).

Download Full Size | PDF

The fractional volumes of both membrane (Fig. 7(A)) and chloroplast (Fig. 7(B)) decreased rapidly as cell sizes increased from ∼3 to 30 µm, reaching a plateau for cells of sizes ∼50 µm. Conversely, fractional volume of cytosol increased rapidly with cell size, reaching a plateau at sizes > ∼50 µm (Fig. 7(C)). The plateau values are approximately 3%, 1%, and 96% for membrane, chloroplast, and cytosol, respectively. A simple model of form ${a / {({b + x} )}}$, where x in micrometer represents equivalent sphere diameter, seems to fit the fractional volumes of membrane and chloroplast well. The details of fitted models are shown in Fig. 7.

The bulk RIs (n_bulk) were calculated from RI measurements (n_i) (Fig. 6) and the volume fraction estimations (V_i) for various cellular structures (Fig. 7) as:

As expected, bulk RIs for phytoplankton decreased with increasing cytosol volume fractions (Fig. 8(A)), consistent with Aas [4], who found that bulk refractive index of phytoplankton is largely determined by their partial water volume. As a matter of fact, the linear regression line shown in Fig. 8 is very similar to the model shown in Fig. 2 of Aas [4]. Given the dominant role of cytosol/water content in determining the bulk RI (Fig. 8(A)) and as cytosol/water content increases with cell size, it is not surprising to see a general trend of decreasing bulk RIs of phytoplankton with increasing cell size (Fig. 8(B)). The coefficient 1.339 in the regression model shown in Fig. 8(B) represents the RI of the background medium, i.e., the bulk RI approaches that of cytosol as the partial volume of cytosol increases towards 100%. A similar trend was also obtained for the internal bulk RI (Fig. 8(C)), which represents the volume-weighted average RI for the non-membrane cellular structures.

 

Fig. 8. The bulk RI calculated for different phytoplankton species as a function of (A) partial volume of cytosol and (B) equivalent sphere diameter (ESD). (C) The bulk RI for non-membrane cellular structures as a function of ESD. The mean and standard error of regression are –0.21 ± 0.004, 1.53 ± 0.003in (A); 0.62 ± 0.04, –1.82 ± 0.30 in (B); and 0.31 ± 0.02, –2.73 ± 0.24 in (C). The symbols are the same as in Fig. 7.

Download Full Size | PDF

The partial volumes of various cellular structures for individual species and the corresponding bulk RIs are summarized in Table 2. The volume measurements show that most diatoms have a cell wall that accounts for 20.5 ± 10% of their total volume. The largest cell wall by percent volume was found for Pseudo-Nitszchia spp., measuring at 44.5 ± 21.3%. Due to their needle-like shape, there was greater uncertainty associated with the estimate of their total cell volume and consequently Pseudo-Nitszchia spp. appeared as outliers when plotted in Fig. 7 (not shown). For the phytoplankton measured in this study, the chloroplast typically accounted for < 10–15% of total volume and cytosol 50–80%. The exception is E. huxleyi, with 20% for chloroplast and only 7% for cytosol.

Table 2. Partial volume measured for different cellular structures and the estimated bulk RI for each phytoplankton species as well as ciliates and flocculate.

View Table | View all tables in this article

4. Discussion

4.1 Comparison of bulk RI with previous studies

In general, the bulk RI values estimated for different organisms agree with values obtained in previous studies (Table 3). Specifically, RI values for diatoms agree well with the Aas [4] estimates based on metabolite composition, however, values were lower than values reported by Poulin et al. [47], Grant et al. [48], and Ackleson and Spinrad [7]. One possible explanation is the size of the cells examined. These three previous studies measured the RIs of relatively small diatoms, T. pseudonana and P. tricornatum, which typically have mean ESD of approximately 3.5 µm and 2.5 µm, respectively [49,50]. For comparison, the two most abundant diatoms examined in this study were Skeletonemia spp. and Chaetoceros spp. which had higher mean ESDs of approximately 10 µm and 18 µm, respectively. Generally, larger organisms contained a higher water volume than smaller organisms (Fig. 7(C)), which in turn would lower their bulk RI (Fig. 8(B)).

Table 3. Comparison of bulk Refractive index (RI) values obtained from different studies.

View Table | View all tables in this article

Bulk RI values for the coccolithophores reported here are higher than values reported in [4,7,48] by approximately 6–12%. This is probably due to coccoliths in our study occupying a greater volume (60–80%) of the cell (Fig. 7(A)). Nevertheless, RI values for the coccolithophores are in general agreement with the values reported in [47,51] given the variability of our measured data (Table 3). Aas [4] reported the lowest bulk RI value for coccolithophores. Aas, however, did not consider the internal structure of these organisms, instead opting to consider the phytoplankton as spheres with a homogenous composition. A sphere with a homogenous composition would scatter less light at angles > 10–20 degrees compared to a coated sphere with a distinct external membrane of higher refractive index [22], i.e., diatoms and coccolithophores. This can directly affect the estimates of RI based on optical inversion using light scattering measurements by, e.g., flow cytometry [7] and backscattering [47,51].

For dinoflagellates, it was surprising to see that our values, which were based on one species Prorocentrum spp. agree well with the estimate of Aas [4]. However, our values are lower than the estimates for different dinoflagellate species, Alexandrium catenella and Ceratium spp. in [25] and C. polykrikoides and N. miliaris in [52]. Few studies have measured RIs of ciliates. The one study by Morel and Ahn [19] reported RIs of 1.399 for ciliates of diameters between 10–20 µm. The mean ESD for ciliates measured in this study was smaller at ∼7 µm. Despite this size difference, our values fall within their measurement uncertainty.

4.2 Volumetric cellular structure and sphere models

Our study revealed three salient cellular structures could be distinguished based on their RI: external membranes, chloroplasts for phytoplankton (or cytoplasm for ciliate), and cytosol. The relative proportion of volume of both the external membrane and chloroplast of phytoplankton were found to decrease with cell size (Fig. 7(A) and (B)), approximately scaling with 1/d, where d represents equivalent spherical diameter. As the total cell volume scales with d³, this signifies that the absolute volumes of the external membrane and chloroplast scale with d². Following Okie [53], the d² scaling of their absolute volumes implies that only the surface area of the membrane and chloroplast increase with size whereas their third dimensions remain constant. To better understand this argument, we fitted a model describing the fractional volume (fV_shell) of the shell of thickness h in a coated sphere of radius r to the data shown in Fig. 7(A). The model is of the form as

Additionally, the fractional volume of cytosol increases with cell size logistically, reaching a maximum fraction of ∼96% (Fig. 7(C)). As the average RI of phytoplankton cells is largely determined by the volume of cytosol, the average RI for both bulk particles (Fig. 8(B)) and internal portion (Fig. 8(C)) decrease with cell size.

All the cells examined in this study contained an external membrane or shell, which displayed distinctive RI that varied according to their composition (Fig. 6(A)): silicate, calcium carbonate, cellulose, and phospholipid (Table 1). This supports the application of coated sphere models to simulate optical properties of phytoplankton, and likely explains their better agreement with observations compared to homogeneous sphere models [24,26,27,29]. Moreover, our results offer further insight into the proper configuration of coated spherical models used to simulate the optical properties of organisms: (1) the RI of the shell should be representative of membrane composition, i.e., silicate, calcium carbonate, cellulose, or phospholipid; (2) the thickness of the shell should remain as a constant; (3) the RI of the core should decrease with cell size. Coated sphere models combining membrane and chloroplasts into the shell and considering the core as solely cytosol have also been used to simulate the optical properties of phytoplankton [25,28]. Although our measurements do not support such a model scheme, the combined volume fractions of membrane and chloroplasts still decrease with cell size, scaling approximately with 1/d (Figure S1 in Supplement 1). Finally, if a homogeneous sphere model must be used, then the bulk RI should decrease with cell size.

5. Conclusions

To our knowledge, this is the first time that the refractive index and the volumetric distributions of live plankton cells have been reported. The 3D Cell Explorer was able to clearly differentiate three separate cellular structures according to their refractive index: membrane, chloroplast for phytoplankton (or cytoplasm for ciliate), and cytosol. The RI values of membranes are mainly determined by their composition (Fig. 6(A)), including silicate for diatoms, calcite carbonate for coccoliths, cellulose for dinoflagellates and naked E. huxleyi, and phospholipid for ciliates. Despite the larger intra-species variability of RI values for chloroplasts compared to those of membranes, no statistically significant differences were found across species (Fig. 6(B)), suggesting the composition of chloroplast membrane is relatively consistent across different species. Unsurprisingly, the RI values for cytosol were similar to that of seawater (Fig. 6(C)).

The relative volume fractions of both membrane and chloroplast decrease with cell size and scale approximately to the inverse of cell size (Fig. 7(A) and (B)). This suggests that the volumes of the two cellular components scale approximately with d². For membranes, this suggests that thickness remains constant while its surface area expands as cell size increases. Conversely, the relative volume fraction of cytosol increases with cell size following a logistic curve (Fig. 7(C)). The fractional volumes of membrane, chloroplast, and cytosol seem to reach plateaus of 3%, 1%, and 96% at sizes > 50 µm. The volume-weighted average refractive indices of the plankton species decrease with both cytosol volume and cell size. Our findings provide valuable insights for the configuration of particle models used for computing the optical properties of plankton. Moreover, our study also demonstrates the potential of the 3D Cell Explorer to obtain difficult surface area-volume scaling information of cells that can lead to new exploration of the fractal geometry of phytoplankton [53].