Label-Free Detection of Lipid Accumulation in Cells Using Magnetic Levitation

Abstract

Dysfunction in adipose tissue can cause serious health problems, including obesity, type-2 diabetes, and cardiovascular disease, significantly reducing human life expectancy. Differences in differentiation and lipid accumulation in adipocytes reflect their functional status, making it important to characterize adipocytes by monitoring biophysical changes during adipogenic differentiation. However, there is currently no specific cell surface marker to separate mature adipocytes from non-adipose cells based on their lipid content, and separation of mature adipocytes is challenging due to handling limitations without fixation, antibody staining, or particle conjugation. Here, we report a biomarker-free, magnetic levitation-based method to detect density changes and quantify the accumulation of lipid-rich droplets within differentiating adipogenic cells. Magnetic levitation revealed density changes within preadipocytes differentiating towards mature adipocytes, with density decreasing over time as cells accumulated lipids. We then used lipid droplets as an intracellular marker to quantify lipid accumulation in single adipocytes during adipogenesis. The significant density changes correlated with cell morphology and lipid droplet morphology within the cytoplasm. For the first time, free-floating lipid vesicle density was measured using magnetic levitation. This unique method enables efficient detection and quantification of dynamically evolving lipid droplets in cells, proving beneficial for modeling lipid storage-related diseases and drug screening applications.

1. Introduction

Obesity is a major health problem, affecting at least 1 billion of the world’s population.^[1,2] It causes an estimated 5 million deaths per year, largely due to medical complications such as diabetes and cardiovascular disease.^[1] Increases in lipid accumulation in the lipid droplets (LDs) in adipose tissue that occur with obesity and results in adipocyte hypertrophy are particularly associated with adverse cardiometabolic outcomes.^[3] Adipose tissue is known for its regulatory responses for balancing energy levels, controlling body temperature, blood glucose levels, and lipid homeostasis, as well as for secretion of multiple hormones and growth factors, and has an active role in influencing inflammatory responses.^[4,5] Outside of obesity, LD accumulation in cells reflects and contributes to rare but highly morbid genetic conditions, such as neutral lipid storage diseases.^[6] Therefore, understanding the function of LDs is essential to study adipocyte metabolism and its contributing molecular factors for the treatment of LD-mediated and obesity-mediated diseases.^[7]

However, adipocytes are highly fragile and deformable under mechanical stresses due to the liquid triglyceride-containing LDs in their cytoplasm, which makes it difficult to handle or isolate efficiently to identify subpopulations to perform single-cell analysis for fat cell development based on their lipid content.^[8,9] For instance, preadipocytes, i.e., fibroblast-like quiescent 3T3-L1 cells; have been extensively used to analyze adipogenic differentiation and lipogenesis process upon differentiation with exogeneous stimuli.^[10] These cells are highly heterogeneous, resulting in a wide range of size, morphology, and lipid accumulation during adipogenesis, due to the dynamically evolving mixture of phenotypically distinct subpopulations.^[11] Additionally, the adipocyte differentiation process is affected by differentiation protocols and inherent cell properties (e.g., passage number).^[12] The asynchrony of the differentiation process and the molecular mechanisms of adipocyte differentiation and LD accumulation are important to understand to develop novel therapeutic strategies for metabolic disorders, such as obesity, diabetes, and cardiovascular diseases.^[5,13]

Currently, there is no specific cell surface marker to separate mature adipocytes from nonadipose cells based on their lipid content, without fixation, antibody staining,^[14] or conjugation of magnetic beads.^[15] Even though flow cytometry has been widely used to separate LD-containing adipocytes based on the refractive index of LDs or on fluorescent labeling for cellular phenotyping including CD34 (preadipocyte marker),^[15] or using LipidTOX dyes^[8]; mature adipocytes or adipocyte clusters are too large (diameter >120 μm) to pass through a conventional cytometer nozzle.^[8] Only a small percentage of mature adipocytes generally survive the sorting process due to high-pressure inflow, smaller fluidic diameters, and increased shear stress.^[8] Alternatively, Raman image-activated cell sorting (RIACS) method^[16] can be used to sort differentiated adipocytes without relying on fixation or staining. This technique resulted in a sorting purity of 74.4%, which included cells that have a large lipid density and a large spatial distribution of LDs. However, it faces some technical limitations in terms of its high cost and requirement for advanced instrumentation.^[16] Another approach that has been proven to be useful for separating cells based on the inherent density was demonstrated using a magnetic levitation-based technology,^[17] which enabled the detection of changes in metabolic activities,^[18] responses the various drug treatments,^[17] separation of diseased cells from healthier ones^[19–22] or immune responses^[19] in real-time depending on the inherent differences in density properties. The separation of cells based on their intrinsic physical properties, such as density, is an attractive method, as it does not require labels or tags compared to the existing methods.^[23] A magnetic levitation system allows for the investigation of cellular changes uniquely enabling ultra-precise density measurements at the single cell level with 0.0001 g mL⁻¹ precision.^[17] The density of cells can be altered by biophysical changes such as alterations of their lipid content. For example, a patient with neutral lipid storage disease resulted in a higher density profile for healthy induced pluripotent stem cells (iPSCs)-derived cardiomyocytes (CMs) compared to the adipose triglyceride lipase (ATGL)-deficient iPSCs-derived CMs.^[22] This method does not rely on the shape or volume of measured samples and does not exert high pressure, shear forces, or other mechanical stresses on the cells, which makes it ideal for the manipulation and characterization of fragile cells.

Here, we present a biomarker-free, magnetic levitation-based method to precisely detect density changes and quantify the accumulation of lipid-rich vacuoles within differentiating preadipocytes (Figure 1). To date, changes in density profiles depending on lipid accumulation have never been demonstrated before for differentiating preadipocyte cells toward an adipocyte lineage. As a model to validate our method, we used mouse 3T3-L1 cells that constitute the most frequently used and well-established pre-adipocyte system, sharing many properties with normal adipocytes.^[24] We demonstrate that levitation and density signatures are significantly altered during adipogenesis. Moreover, we use LDs as an intracellular marker to quantify lipid accumulation during adipogenesis. We envision that this inexpensive, portable, and versatile biophysical characterization method can be broadly applicable to separate populations based on their lipid content driven by differential density properties at the single-cell level.

Figure 1.

Label-free profiling and characterization of lipid accumulation in cells using magnetic levitation. A1) As a model system, 3T3-L1 preadipocytes were differentiated over 12 days, which results in intracellular lipid accumulation, leading to changes in levitation heights and density (Dexa: Dexamethasone, and IBMX: 3-Isobutyl-1 methylxanthine). A2) Schematic of the levitated cells under the magnetic field during the differentiation process and their separation based on their inherent density profiles. Sub-populations with different densities form distinct levitation zones due to differential lipid accumulation, which can be further quantitatively profiled at different levitation heights (F_magnetic, F_buoyancy, and F_gravity represent the magnetic, buoyancy and gravitational forces during magnetic levitation, respectively). (Created with Biorender.com).

2. Results and Discussion

2.1. Magnetic Levitation of Adipocyte-like Cells During Differentiation

Here, we apply the principles of magnetic levitation to levitate and separate lipid-accumulating cells based on their inherent densities, without relying on any biomarkers, antibodies, or tags.^[17] The magnetic levitation device (MagDense) consists of a capillary channel sandwiched between two permanent magnets (NdFeB) with the same poles facing each other.^[17] Cells are suspended in a nontoxic, nonionic paramagnetic medium. The magnetic susceptibility difference between a differentiating, lipid-accumulating cell and the surrounding paramagnetic environment causes it to move away from a higher to a lower magnetic field gradient. Cells are then focused and levitated at a certain height within the capillary, where the magnetic ($\overrightarrow{F_{mag}}$), buoyancy ($\overrightarrow{F_b}$), and gravitational ($\overrightarrow{F_g}$) forces reach an equilibrium (Figure 1A2). As we have previously demonstrated, the final levitation height of a cell mainly depends on its inherent density.^[20] Biophysical changes during differentiation can alter the size, shape, or mass of individual cells, which can potentially change the density of their sub-cellular components resulting in an overall change in cellular density.^[25] We further applied magnetic levitation profiling to specifically rank and separate differentiating adipocyte-like cells within seemingly homogenous populations, according to their lipid content.

To assess the capabilities of the magnetic levitation platform to detect lipid accumulation in cells during adipogenesis, 3T3-L1 preadipocytes were differentiated for 12 days. Differentiating preadipocytes were collected on days 3, 6, 8, and 12 from culture plates, and then levitated by suspending the heterogonous populations to characterize their biophysical properties. The magnetic levitation platform was able to detect differentiating 3T3-L1 cells committing to adipogenesis without relying on labels or markers, based on inherent density changes (Figure 2A1–5). The corresponding levitation height profiles on Days 0, 3, 6, 8, and 12 are shown in Figure 2B1–5. Undifferentiated 3T3-L1 preadipocyte cells reached their equilibrium height at 248.14 ± 6.62 μm at day 0 (Figure 2B1). As cells started to differentiate, the mean levitation height significantly increased to 274.72 ± 2.69 μm for Day 3 (p = 0.004), 305.11 ± 3.58 μm for Day 6 (p = 0.0001), 325.82 ± 14.32 μm for Day 8 (p = 0.004), and 372.32 ± 12.25 μm for Day 12 (p = 0.0001), respectively (Figure S2A1, Supporting Information). Although we started to observe LD formation within differentiating cells, this did not lead to a significant change in the distribution of levitation height profiles up to Day 6. Interestingly, with the continuing accumulation of LDs during adipogenic differentiation, the levitation height distribution significantly changed after Day 6. This resulted in ≈1.37- and 1.94-fold increase on Day 8 (p = 0.0001) and Day 12 (p = 0.0001), compared to Day 6, respectively (Figure S2A2, Supporting Information). In addition, free-floating lipid vesicles had their own inherent levitation signatures, which were detected near the top of the capillary (Figure 2A5). Adipocytes are known to be fragile and deformable under mechanical stresses due to their liquid triglyceride-containing droplets in the cytoplasm.^[8,9] Thus, free-floating lipid vesicles could be caused by shear-induced adipocyte lysis exceeding minimal shear stress on cells.^[8] Further, levitation height distribution was more heterogeneous on Day 12 (≈1.42-fold, p = 0.0001), compared to Day 8. Increases in lipid content within the cytoplasm could lead to hypertrophy,^[26] and this might cause the cells with the most lipid droplet accumulation to be positioned at higher levitation positions, compared to the rest of the less adipogenic cells on Days 8 and 12.

Figure 2.

Magnetic levitation profiles of adipocytes during differentiation. Heterogeneous sub-populations of adipocytes with different levels of lipid accumulation can be profiled and detected in the levitation system. A1–5) Representative images of levitating undifferentiated 3T3-L1 preadipocytes (Day 0), after treatment with differentiation medium for 3 days (Day 3) and during treatment with maintenance medium on Days 6, 8 and 12 (Day 6, Day 8, Day 12). Arrows indicate LDs and adipocyte clusters. B1–5) Corresponding levitation height profiles of differentiating cells on Days 0, 3, 6, 8, and 12.

2.2. Biophysical Profiling of Levitating Preadipocytes, Adipocytes and Lipid Vesicles

Next, we characterized the cellular heterogeneity in physical properties, such as density and cell size during differentiation. The undifferentiated 3T3-L1 cells had a mean diameter of 18.28 ± 2.84 (μm) on Day 0. Cell diameter significantly increased over time during adipogenesis, resulting in 20.02 ± 3.88 (μm) for Day 3 (p = 0.016), 20.90 ± 4.42 (μm) for Day 6 (p = 0.0001), 21.54 ± 4.70 (μm) for Day 8 (p = 0.0001), and 22.26 ± 6.21 (μm) for Day 12 (p = 0.0001), compared to Day 0 (Figure 3A; Figure S3A1–5, Supporting Information). Our imaging algorithms detected similar size ranges for preadipocytes and adipocytes, comparable to those measured by commercially available systems^[27]. The largest cell diameter was detected on Day 12, which was ≈50 μm for a single differentiated, lipid-accumulating cell. This was similar to the reported size ranges of mature adipocytes.^[28] Next, we quantified the density distributions of undifferentiated and differentiating cells over 12 days. The average inherent density of undifferentiated 3T3-L1 cells was 1.093 ± 0.005 (g mL⁻¹) on Day 0. As cells differentiated and started to store lipids, their inherent density significantly decreased (p = 0.0001), to 1.081 ± 0.010 (g mL⁻¹), 1.073 ± 0.015 (g mL⁻¹), 1.071 ± 0.017 (g mL⁻¹) and 1.059 ± 0.027 (g mL⁻¹) on Days 3, 6, 8 and 12, respectively (Figure 3B). The density of native adipose tissue in humans was reported to range from 0.925 to 0.970 g mL⁻¹, depending on the lipid, water, and dry fat-free component.^[29] Next, we analyzed the heterogeneity of individual cell densities (Figure S3B1–5, Supporting Information). The normalized frequency distribution of single-cell density demonstrated a shift toward lower density values over the course of differentiation. In addition, we showed that adipogenesis not only causes a decrease in single-cell densities but also alters the size distribution of individual cells (Figure S4A1–5, Supporting Information). Based on the confidence ellipse at 95% covariance in cell density and diameter, there were few cells clustered out of the fitted ellipses. These mainly consisted of free-floating lipid vesicles due to their smaller diameter (<12 μm) and lower density profile (<0.98 g mL⁻¹). Additionally, the dead or dying cells were not included in the ellipse, due to their high cell density (1.12–1.16 g mL⁻¹) (Figure 3C–G).^[17,30]

Figure 3.

Density profiling of levitating pre-adipocytes, adipocytes,^[1] and lipid vesicles. A) The corresponding cell diameter; and B) single-cell density profiles of the levitated undifferentiated pre-adipocytes on Day 0 (n = 136) and differentiating pre-adipocytes on Day 3 (n = 129), 6 (n = 167), 8 (n = 90) and 12 (n = 137). Data for 3A and 3B were presented as mean ± st.dev. (*: compared to Day 0, ^***p < 0.001, ^**p < 0.01, *p < 0.05). (#: compared to Day 6, ^###p < 0.001, ^##p < 0.01, ^#p < 0.05). (&: compared to Day8, ^&&&p < 0.001, ^&&p < 0.01, ^&p < 0.05). ($: compared to Day 12, ^$ $ $p < 0.001, ^$ $p < 0.01, ^$p < 0.05). C–G) The relationship between cell diameter and single-cell density of levitated cells over 12 days and their corresponding Pearson correlation coefficient (R) values. The data were fitted with confidence ellipse at 95% covariance on Days 0, 3, 6, 8, and 12. H) Pearson correlation coefficient (R) between the cell diameter and density of levitated cells for all time point groups. Bars show all statistically significant comparisons (^***p < 0.001, ^**p < 0.01, *p < 0.05).

The Pearson correlation coefficient (R) values between cell diameter and density changed significantly over time (one-way ANOVA, p = 0.0004) (Figure 3H; Table S1, Supporting Information). On Day 0, undifferentiated 3T3-L1 cells were correlated negatively, resulting in overall R = −0.1398 ± 0.14 in readouts at the population level. After treatment with the differentiation media, the correlation significantly changed, resulting in a positive value of R = 0.117 ± 0.09 on Day 3 (p = 0.046). As cells continued to accumulate more lipids, the correlation changed its direction and became negative again, R = −0.096 ± 0.07 on Day 6, R = −0.29 ± 0.14 on Day 8 and R = −0.23 ± 0.16 on Day 12, where an increase in lipid content and cell diameter resulted in a decrease in density. Even though the Pearson correlations started lower at the beginning of the differentiation process, correlations increased over time, which was significantly different on Day 8 (p = 0.0003) and Day 12 (p = 0.0015), compared to Day 3. These results align with our description of the biological changes the cells are undergoing during the differentiation process as shown in Figure 3A,B. Immediately after treatment with the differentiation media, the sudden start of the differentiation process led to the positive change in R; however, over time, as the cells accumulate lipids and the cell size increases, the density of the cells decreases leading to a negative change in R. These biological factors help explain why Day 3 is a turning point in the Pearson correlation and why the comparisons of Day 3 with other time points are statistically significant. We acknowledge that the R values indicate modest correlations. The fact that the strongest correlation explains only ≈15% of the variance reflects the complex, multifactorial nature of adipocyte differentiation. The remaining variance likely reflects the heterogeneous nature of adipocyte differentiation, including factors such as cell cycle state, local microenvironment effects, variations in individual cell responses to differentiation stimuli, and inherent biological variability due to asynchronous differentiation within the differentiating adipocyte population.

2.3. Formation of Distinct Levitation Zones based on Lipid Accumulation

Based on the single-cell density measurements in the magnetic levitation system, we separated the levitated cells into four distinct zones (Figure 4A). The levitation zone I represents the cells with an inherent density in the range of ≥1.12–1.16 g mL⁻¹. This zone might form due to the cell debris, and dead or dying cells.^[22,32] This density range aligns with previously reported values for dead/dying cells in our earlier work^[17,30] Microscopic examination showed these cells/particles in Zone I lacked typical cellular morphology, and these particles showed no lipid droplet staining with either BODIPY or Nile Red. The levitation zone II represents the cells that have a density of ≥1.06–1.12 g mL⁻¹. This zone includes the undifferentiated 3T3-L1 cells or less adipogenic cells with no distinguishable separation, based on their levitation profiles. Although these cells were committed to adipogenic differentiation, they did not accumulate enough LDs to alter their density signatures significantly. Levitation zone III represents the cells that have a single cell density of ≥0.98–1.06 g mL⁻¹. These include the mature adipocytes, that accumulated enough LDs to be distinguished from the undifferentiated cells. Due to these significant density differences, we observed a clearer separation of mature adipocyte cells in zone III from the undifferentiated cells, which could enable their sorting with high purity for further downstream analyses. Lastly, the levitation zone IV mainly consisted of free-floating lipid vesicles at the density range of 0.93 to 0.98 g mL⁻¹. Previously, bone marrow-derived stem cells were differentiated toward adipogenic cell lineage, and then their density profiles were characterized using magnetic levitation, where differentiated and lipid accumulated cells were located with density of lesser than 1.02 g mL⁻¹ and undifferentiated cells were located with a density of greater than 1.02 g mL⁻¹.^[9] In a recent study, adipogenic differentiation of 7F2 cells was levitated on Day 15 and assessed using a HologLev system. However, the characterized holographic images did not show a mean density difference between undifferentiated control cells (Day 0) and differentiated 7F2 cells on Day 15.^[31] In our study, we demonstrated the levitation profiles of differentiating preadipocytes on days 3, 6, 8, and 12, which were identified in multiple density regions including undifferentiated cells, differentiated preadipocytes but not accumulated enough LDs to separate themselves from the undifferentiated population, mature adipocytes, and the free-floating lipid vesicles. Furthermore, we demonstrated the morphological assessment of individual lipid droplets within cell cytoplasm and its effect on the overall density change, which was not shown previously.

Figure 4.

Formation of distinct magnetic levitation zones based on lipid accumulation. A) The separation of cells based on the inherent density differences in levitation zone I (1.12–1.16 g mL⁻¹), zone II (1.06–1.12 g mL⁻¹), zone III (0.98–1.06 g mL⁻¹) and zone IV (0.93–0.98 g mL⁻¹). B) The separation percentage of levitated cells on Days 0 (n = 4), 3 (n = 5), 6 (n = 5), 8 (n = 6), and 12 (n = 6) depending on the levitation zones (data were presented as mean ± s.e.m.). C) The density and D) diameter of free-floating lipid vesicles (n = 18) in levitation zone IV (data were presented as mean ± st.dev.).

Next, we investigated how these different levitation zones contribute to the separation of differentiating cells over time (Figure 4B). It is crucial to detect the cells committing to an adipogenic lineage at the earliest stages of differentiation. According to the levitational imaging, the beginning of observable separation due to significant lipid accumulation was detectable on Day 3, in levitation zone III. However, the differentiation and lipogenesis process did not cause a significant separation of mature adipocytes until Day 8. For instance, separation percentage based on the lipid accumulation resulted in a ≈3.6-fold increase in zones III and IV on Day 8, compared to Day 6 (p = 0.001). We observed the most significant increase, 25.6-fold, in separation percentage based on lipid accumulation on Day 12, compared to Day 3 (Figure S2A3, Supporting Information). There were no cells detected in zones III and IV on Day 0 (Figure 4B). As 3T3-L1 cells were differentiating, the number of cells in zone III significantly increased while cell numbers in zone II significantly decreased. This resulted in a ≈26.9-fold increase in zone III on Day 12, compared to Day 0. In addition, free-floating lipid vesicles were only observed in zone IV. Next, for the first time, we characterized the mean densities of free-floating lipid vesicles in the magnetic levitation system. The density these free-floating lipid vesicles was 0.955 ± 0.013 g mL⁻¹ (Figure 4C) and their average diameter was 8.83 ± 2.08 (μm) (Figure 4D). The lipid fraction was previously recorded to be 0.54–0.85 g/mL from an adipose tissue^[29]. The differentiated cells started to form large clusters at Day 12 since the separation percentage (%) due to lipid accumulation significantly increased over time in zone III.

2.4. Lipid Droplet Analysis within Levitated Cells Based on Their Density Profiles

The adipogenic differentiation of 3T3-L1 cells has been shown to be highly heterogeneous in the literature.^[11] Similarly, we observed that accumulated LDs within individually differentiating 3T3-L1 cells were visibly heterogeneous. To better understand and quantify the fat cell differentiation and maturation within the differentiating 3T3-L1 cells, we investigated the cellular LDs accumulation (Figure 5). To investigate the density changes based on lipid accumulation, we mixed cells from Day 0, 6, and 12 levitating after performing LD staining (Figure 5A). As expected, lipid accumulating cells on Day 6 (green) and Day 12 (red) formed levitation bands, above the undifferentiated cells from Day 0. The normalized density (%) with respect to Day 0 significantly differed compared to Day 6 and 12 (p = 0.0001), where density is significantly decreased over time as a response to the progression of differentiation (Figure 5B). In addition, the most significant separation percentage among the lipid-containing cells was measured on Day 12 (p = 0.0001), resulting in 66.8 ± 6.21% (Figure 5C). In addition, we showed that as individual cells accumulated more LDs, they became significantly less dense, and single-cell density distribution among the lipid-accumulated cells were shifted toward to a lower density gradient (Figure 5D). The total LD area with respect to cell area with relation to cell density was plotted according to top (<1.06 g mL⁻¹) and bottom (≥1.06 g mL⁻¹) regions, where the confidence ellipse at 95% covariance were fitted on data (Figure 5E). Even though a majority of the cells on Day 12 were in the top region, there were a few cells detected in the bottom region, which had a density lower than 1.06 g mL⁻¹ and had a higher percentage of accumulated LDs in the cell cytoplasm compared to Day 6.

Figure 5.

Lipid droplet analysis within intact, levitated cells based on their density profiles. A) Levitated cells were analyzed for lipid vesicles, where a cell mixture including undifferentiated 3T3-L1 cells from Day 0 (unstained), differentiating cells on Day 6 (BODIPY-green) and on Day 12 (Nile Red-red) suspended within the magnetic levitation system. B) The percentage of single cell density of levitated cells on Day 6 (n = 167) and 12 (n = 137) normalized with respect to Day 0 (n = 137). C) The separation percentage (%) of LDs within intact cells on Day 6 (n = 11) and 12 (n = 11). D) Normalized frequency distribution of LDs within intact cells at various density ranges. E) Scatter plot represents the relationship between the total lipid area divided by cell area (%) versus single cell density on Days 6 and 12. Principal component analysis was applied to present the correlation circles of each subpopulation. Confidence ellipse at 95% covariance fitted based on the top and bottom regions. Data were presented as mean ± st.dev. *p < 0.05, ^**p < 0.01, ^***p < 0.001.

Next, we characterized the LDs within intact, levitated cells on Days 6 and 12 (Figure 6). We detected an increase in the number of mature adipocytes on Day 12, compared to Day 6. Representative images of individual LDs detected within differentiating adipocytes are shown in Figure 6A1–2. The total number of LDs in levitating cells was significantly increased by ≈2.09-fold on Day 12, compared to Day 6 (p = 0.0001) (Figure 6B). The area of single LDs also significantly increased on Day 12 (p = 0.048), compared to Day 6 (Figure 6C). As a result, there was a ≈2.4-fold increase in the total lipid area versus cell area on Day 12 (p = 0.0001), (Figure 6D). This could be due to the addition of insulin in the adipocyte maintenance medium, which accelerates lipid accumulation and increases the size of LDs within differentiating preadipocytes.^[12] We also analyzed the circularity of the individual LDs within cells on Days 6 and 12. Interestingly, the circularity of LDs decreased over time (p = 0.003) (Figure 6E) and the Feret diameter of single LD increased on Day 12 (p = 0.001), compared to Day 6 (Figure 6F). Overall, the decrease in LD circularity and increase in single LD area resulted in a decrease in overall cell density on Day 12, compared to Day 6, due to adipogenic differentiation (Figure 6G1–2). Additionally, increases in Feret diameter of LDs led to a decrease in cell density, as a response to increases in lipid accumulation area within the cytoplasm (Figure 6H1–2).

Figure 6.

Characterization of lipid droplets within intact, levitated cells after 6 and 12 days of differentiation. A1–2) The single lipid droplet (SLD) analysis on the levitated cells on Day 6 (green) and Day 12 (red). B) The total number of LDs per single cell on Day 6 (n = 31) and 12 (n = 32). C) SLD area (μm²) on Day 6 (n = 122) and 12 (n = 225). D) Total lipid area divided by single cell area (%) on Day 6 (n = 31) and 12 (n = 32). E) Circularity of SLD within cells on Day 6 (n = 122) and 12 (n = 225). F) Feret diameter (μm) of SLD within intact cells on Day 6 (n = 122) and 12 (n = 225). G1–2) The 3D cluster formation based on the relationship between single-cell density (g mL⁻¹), circularity of LDs, and SLD area (μm²) on Day 6 (#clusters: 31, #elements in biggest formed cluster: 37 and the percentage (%) with respect to overall population: 30.3%) and Day 12 (#clusters: 54, #elements in biggest formed cluster: 26 and the percentage (%) with respect to overall population: 11.56%). H1–2) The 3D cluster formation based on the relationship between single-cell density (g/mL), Feret diameter (μm) of LDs and total LD area per single cell area (%) on Day 6 (#clusters: 25, #elements in biggest formed cluster: 39 and the percentage (%) with respect to the overall population: 31.97%) and Day 12 (#clusters: 24, #elements in biggest formed cluster: 60 and the percentage (%) with respect to overall population: 26.7%). (Data were presented as mean ± st.dev., *p < 0.05, ^**p < 0.01, ^***p < 0.001).

3. Conclusion

Here, we demonstrated the feasibility of measuring density changes based on lipid accumulation in single cells, using magnetic levitation. Our magnetic levitation-based method uniquely identified cells committing to adipogenesis, from heterogeneous populations of differentiating preadipocytes. In addition, the magnetic levitation platform enabled for monitoring and quantification of dynamically evolving LDs within adipocytes. Determination of lipid accumulation based on very small density changes can help to detect mature adipocytes as a function of their density, cellular content, and differentiation states. This could also be used for single-cell sorting and profiling to investigate adipogenesis-related disorders, to develop novel therapeutic strategies and diagnostic methods.^[32] Furthermore, magnetic levitation methods could be used to identify the differentiated cells depending on LD accumulation for drug screening, which can further expand the possibilities for personalized medicine for lipid storage diseases. The current platform runs under static (non-flow) conditions to profile the biophysical properties of lipid-accumulating cells. The magnetic levitation device can be integrated with microfluidics to isolate cells of interest levitated at different levitation zones for further downstream analysis. In addition, this method is broadly applicable to sort various large and fragile cell types, which generally do not survive the sorting process with the traditional flow cytometry technologies (i.e., Fluorescence-activated cell sorter (FACS).^[33,34] These include pancreatic islets (50–500 μm),^[35] trophoblasts (200–300 μm),^[36] and cardiomyocytes (125 × 25 μm),^[22,33] model organisms such as Caenorhabditis elegans (C. elegans) (≈1.15 mm),^[37] primary human organoids and tumor spheroids. Thus, magnetic levitation-based cellular analysis platforms can broadly enable fundamental research and overcome the current technical limitations related to the handling and sorting of fragile and large cell types.

4. Experimental Section

Cell Culture:

3T3-L1 (ATCC CL-173) mouse embryonic fibroblast cells were plated at a density of 8 × 10⁴ cells per well on 6-well plates in expansion medium, which consisted of 90% Dulbecco’s Modified Eagle’s Medium (DMEM; Gibco) and 10% fetal bovine serum (FBS) (JR Scientific Inc.). Cells were incubated at 37 °C in a humidified 5% CO₂ incubator. Two days post confluency, the expansion media was changed to the differentiation medium, which consisted of 90% DMEM, 10% FBS, 1.0 μm Dexamethasone (Dexa; Sigma-Aldrich), 0.5 mm 3-Isobutyl-1methylxanthine (IBMX; Sigma-Aldrich) and 1.0 μg mL⁻¹ Insulin (Sigma-Aldrich). This was determined as “Day 0” for the adipogenic differentiation process. After 48 h, the media was changed to the adipocyte maintenance medium, which consisted of 90% DMEM, 10% FBS and 1.0 μg mL⁻¹ Insulin. The cells were then differentiated for up to 12 days. Culture media was changed every two to three days. Prior to magnetic levitation at each time point, the culture was treated with 0.25% Trypsin/EDTA solution (Sigma-Aldrich) for 5 min at 37 °C. Then, collected cell suspensions were centrifuged at 300 ×g for 5 min at room temperature before levitation experiments.

Magnetic Levitation System Setup and Calibration:

The magnetic levitation platform was developed in-house, based on the previous work.^[17] Briefly, the device consisted of two coaxially aligned permanent N52-grade neodymium magnets (NdFeB) with same polar surfaces facing each other. A capillary channel with 1 × 1 mm cross-section was placed between the magnets for the levitation experiments. The levitation device was calibrated using commercially available fluorescent polyethylene microspheres (Cospheric LLC) with various densities (1.025 to 1.089 g mL⁻¹) prior to experiments. The microspheres were suspended within FDA-approved, gadolinium-based paramagnetic media (i.e., Gadavist) diluted in PBS at a final concentration of 30 mm. This concentration was shown to be the optimal levitation condition for cells in the previous studies.^[17,20,22] The standard curve was generated based on the relationship between the levitation height and density.

Magnetic Levitation of Adipocytes:

The cells on days 0, 3, 6, 8, and 12 of differentiation were suspended in a culture medium with 30 mm paramagnetic medium. A 30 μL sample was loaded into the capillary and levitated. The cells reached their equilibrium height within 10 min. Levitated cells were imaged under a bright field microscope under 5X or 20X objective.

Characterization of Levitated Preadipocytes and Adipocytes:

The corresponding levitation height profiles and diameter of single cells were characterized using an in-house written MATLAB script (Figure S1A1, Supporting Information). The levitation height distribution was calculated based on the total area under the levitation height profile curves. The density of individual levitated cells and free-floating lipid vesicles was determined according to the generated standard curve, based on the levitation height versus density plots (Figure S1A2, Supporting Information). The separation percentage (%) of levitated cells was determined by dividing the number of cells that have a density of lower than 1.06 g mL⁻¹ by the total number of levitated cells per each trial and multiplying by 100. The threshold of 1.06 g mL⁻¹ was determined based on running a student t-test between the categorized density levels. For the analysis of single adipocytes, a covariance confidence ellipse was applied to determine the concentration ellipses at 0.95 confidence level between levitation height and cell diameter using Chi-Square distribution.^[38] The Pearson correlation coefficient (R) was determined using the linear curve fit as the slope of cell density and diameter curve.

Hierarchical Cluster Formation:

Sub-clusters for cell density and diameter were determined based on the point-to-point distance using the “clusterdata” function in Matlab.^[39] To bring the data to the same scale distance, the cell density and diameter were normalized according to its own range. The distance was specified by determining the optical cutoff threshold versus cluster number, similar to a previously published study,^[40] where the cutoff value increases as the number of clusters decreases to a single cluster.

Visualization of Lipid Vesicles within Levitated Cells:

Differentiating 3T3-L1 cells were stained on Day 6 with BODIPY 493/503 nm (Thermofisher) and Day 12 with Nile Red (AAT Bioquest) to visualize the accumulated LDs within adipocytes. Images of stained and levitated samples were taken under 5X and 20X objectives at Alexa Fluor 488 and 565 nm for Days 6 and 12 samples, respectively.

Characterization of Free-floating Lipid Vesicles:

The free-floating lipid vesicles were visualized using BODIPY stained images on Day 6, as explained previously. The levitation height and density measurements were performed using an in-house written MATLAB script at 0.90 sensitivity. The minimum threshold for size determination was set to be 2 μm to avoid the noise signal. A total of seven images were used to measure free-floating lipid vesicles within the capillary channel inside the MagLev device.

Characterization of Lipid Vesicles:

The morphological characterization of BODIPY- and Nile Red-stained LDs was performed using MRI_Lipid Droplets plugin tool (http://dev.mri.cnrs.fr/projects/imagej-macros/wiki/Lipid_Droplets_Tool) in the ImageJ Software (NIH). The high-resolution fluorescent images were inserted in ImageJ. First, the images were converted into 8-bit grayscale, and brightness and contrast was modified to visualize the LDs within intact cells. The “despeckle” filter was applied to remove the noise signal. Next, using the “Remove Outliers” tool, a set radius of 2.0 pixels and a threshold value of 50 were used to clear the noise further. Thereafter, the minimum size of the object threshold was set up to be “0”. From the LD analysis results, the individual LD area and circularity were determined. LD areas that are less than 10 μm2 were excluded from the analysis. The circularity range was determined from 0 to 1; where ‘1′ stands for the perfect circle and ‘0′ stands for the elongated polygon.^[41] The total area of LD per single cell was calculated by summing up the area of each LD for Days 6 and 12. Similarly, the number of LDs per single cell was calculated by summing up the total number of LDs in each cell for Days 6 and 12, respectively. The “Analyze Particles” tool in ImageJ was used to measure the single cell area (μm2) after determining the region of interest (ROI) by the polygon selection tool. Then, the total area occupied by LDs per cell was calculated by dividing the total LD area by cell area and multiplying by 100. The averaged Feret diameter of LD was calculated according to Equation (1).^[42]

$$<D_f\ge \left(\frac{\mathit{Perimeter}}{\pi }\right)$$ (1)

The separation percentage (%), a covariance confidence ellipse, and the LD containing cell density were calculated as described above. The area (μm²) and number of LDs were measured and counted from a total of 32 cells for each time point. While data was collected from many cells, 32 cells were selected per time point for detailed lipid droplet analysis to ensure consistent image quality and complete cellular boundaries were visible for accurate measurements. This sample size provided sufficient statistical power for our analyses while maintaining stringent quality control. To form hierarchical clusters between single-cell density, circularity, and single LD area as well singl-cell density, averaged feret diameter of LD and total LD area per total cell area, the written “clusterData” code in Matlab was used at the default sensitivity value.^[43]

Statistical Analysis:

All data were presented as mean ± standard deviation (s.dev.), unless stated otherwise. The coefficient of Variation (CoV) was used as the standard deviation divided by mean to reflect the variability within and between experiments. To compare the two groups, the data were analyzed using a two-tailed student’s t-test analysis with a 95% confidence interval at the assumption of equal variances. To compare three or more groups, data were analyzed by one-way analysis of variance (ANOVA) with post hoc comparisons using Tukey’s tests for differences of means at a 95% confidence level. Differences were considered significant at *p < 0.05, ^**p < 0.01, and ^***p < 0.001. All statistical analysis was performed using Minitab 17.3 (Minitab Inc.) or Prism 10 (GraphPad LLC).

Supplementary Material

Supporting Information

Supporting Information is available from the Wiley Online Library or from the author.

Data Availability Statement

Research data are not shared.