Compliance of myeloid cells is lineage-specific

To measure the compliance of cells that are normally in suspension in their physiological environment, a method that preserves this suspended state is required. Here, we used a microfluidic optical stretcher (OS), which is a specific dual-beam trap, capable of inducing well-defined stresses on whole single cells in suspension and therefore useful for measuring overall cell deformability and creep compliance [16], [17]. For details, see the Methods section. Briefly, the OS consists of two diverging, counter-propagating Gaussian laser beams, which emanate from single-mode optical fibres (Fig. 1A). During a typical experiment, cells are introduced into a microfluidic delivery system, serially trapped and then stretched along the laser beam axis (Fig. 1B). The time-dependent strain is extracted from the video camera images, normalized by the peak stress applied and a geometric factor [18], to obtain the creep compliance for each cell. For the cell shown, the peak compliance (at t = 4 s) is 0.083±0.005 Pa⁻¹.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Creep compliance test distinguishes between cell lineages following differentiation.

A, Schematic of the optical stretcher (OS), showing two diverging, counter-propagating laser beams emanating from single-mode optical fibres. B, A cell is held at trapping stress of 1.7±0.1 Pa at time t = 0 s. The trapped cell is stretched when the optical stress is increased to 6.0±0.1 Pa, producing an axial strain of 8.17±0.05% at t = 4 s, for the cell shown. The graph shows both the strain and compliance profiles for this one cell, illustrating the single cell resolution of the OS method. Scale bar is 10 µm. C, Representative compliance profiles for all three fully differentiated lineages and for the undifferentiated cells (n = 89). For neutrophils (Neu; n = 75) and monocytes (Mono; n = 85), the compliance increased, while for macrophages (Mac; n = 52) the compliance decreased. D, Box plots of the average compliance during the four seconds of stretching. For all three differentiated lineages, the entire compliance profile, as well as the peak compliance at 4 s (not shown), were significantly different from undifferentiated HL60 cells, where **** and ** indicate significant differences with p values of <0.0001 and <0.01, respectively.

https://doi.org/10.1371/journal.pone.0045237.g001

The precursor (HL60) cells are induced to differentiate using standard protocols (see Methods and Methods S1) and then repeatedly measured with the microfluidic OS over the course of several days to track the evolution of their viscoelastic properties. Mature neutrophil- and monocyte-like cells are obtained from HL60 cells 72–96 hours after induction of differentiation. Macrophage-like cells are obtained from 24 hours after induction of differentiation. Our results show that a creep compliance measurement distinguishes between these three cell lineages following differentiation to mature cells (Fig. 1C, D). For all three lineages, the peak compliance at t = 4 s of stretching and the entire compliance profile are significantly different from undifferentiated cells (p<0.0001). Moreover, even though both neutrophils and monocytes become more compliant, we still find significant differences between both their peak compliances and their entire compliance profiles (p<0.01). These lineage-specific differences in compliance are reproducible at higher stresses and in different OS set-ups with varying geometries (Fig. S2). Of note, concomitant changes in cell size during differentiation (Fig. S3) are explicitly considered in the geometric factor F_g, which is used for the proper normalization of the compliance. Thus, the creep compliance is a true material property of these cells and can be used as sensitive yet robust, non-invasive and inherent physical marker to monitor differentiation and even lineage specification.

Lineage-specificity is reflected in detailed viscoelastic properties

To gain new insights into cell biological function using cell material properties further analysis of the compliance results is required. We employed both basic mechanical and power law models to extract detailed viscoelastic parameters (Fig. 2A; see also Fig. S4 and Methods S1). The equation for the power law model that fitted the data best required an offset A,(1)where J(t) is the time-dependent compliance, J_o is the initial compliance, τ₀ = 1 s and β is the power law exponent. The lineage specific modulation of compliance found in the raw compliance curves (Fig. 1C) is manifested in the power law fits in the initial compliance, J_o. Tracking both parameters over the course of several days (Fig. 2B), this parameter J_o confirms that all macrophage-like cells (24–72 h after induction of differentiation) are significantly less compliant than undifferentiated cells while neutrophil- and monocyte-like cells become significantly more compliant (p<0.0001). Moreover, neutrophils are more compliant than monocytes at 96 h (p<0.0001). In the power law model, a purely elastic solid has a power law exponent of 0 and a purely viscous fluid has a power law exponent of 1. For macrophages the exponent β increases from 0.65 to about 1 within the first 24 h showing increasing fluid-like behaviour, while it decreases for neutrophils to 0.35 right before they become fully differentiated at 96 h showing progressively solid-like behaviour compared to the undifferentiated cells (Fig. 2B). The exponent β is nearly constant for monocytes, distinguishing them from neutrophils.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Power law and mechanical models describe fate-specific viscoelastic properties of cells during the course of differentiation.

A, Power law fit (P) and one of the mechanical models, here Burgers' model fit, (B) used to extract viscoelastic parameters. For time-scales above 0.1 s and much longer, both power law and mechanical models offer overlapping insights. B, Evolution of initial compliance, J_o, and power law exponent, β, during the differentiation of cells over several days. J_o characterizes the lineage-specific changes in compliance as cells become fully differentiated at 24 h (macrophages) and 96 h (monocytes and neutrophils) after induction of differentiation. The power law exponent β is about 1 for macrophages, indicating liquid-like behavior. The exponent decreases for neutrophils showing increased solid-like behaviour as cells become fully differentiated at 96 h, but stays constant for monocytes, apparently consistent with their state as immature macrophages. C, Time-scales, at which cells change their viscoelastic behaviour, defined as the ratio η₂/E₂ from the transient part of viscoelastic models. D, All differentiated lineages have a lower steady-state viscosity. E, Macrophages are less compliant than undifferentiated HL60 cells on short time creep (4 s) and more compliant on longer creep test (20 s). The * and ** represent significant differences with p values<0.05 and <0.01, respectively.

https://doi.org/10.1371/journal.pone.0045237.g002

Since power law models assume time scale independence, which may not generally hold for biological cells, we also used specific viscoelastic models for fitting to the compliance data. The best-fitting viscoelastic model for undifferentiated HL60 cells, and differentiated neutrophils and monocytes was the so-called Burgers' model (B),(2)while the best-fitting model for macrophages was the standard linear liquid model,(3)where E₁ describes the instantaneous elastic response found in all cell types except macrophages. The mere fact that macrophages are mechanically best represented by a liquid model, while the other cells have a more solid character, further emphasizes the emerging differences. The last term in equations (2) and (3) determines the transient response, which was found in all cell types. Clearly, it is possible to obtain the steady-state viscosity, η₁, when the transient response becomes constant. For comparisons, we focused on the fit parameters found in all cell types, namely, the transient and steady-state parameters, and found the characteristic time-scales, τ, at which cells change their viscoelastic behaviour (Fig. 2C and Fig. 2D, respectively). Here, τ is given by the ratio of transient viscosity, η₂, and elasticity, E₂. The undifferentiated cells had a transition time of τ≈7 s. For monocytes and neutrophils, τ was 5 s but not significantly different from undifferentiated cells. For macrophages it was over 15 s, and was different from all other cell types (p<0.05) reflecting a reduced elasticity, E₂. Regarding the steady-state viscosity, η₁, all three differentiated cell types showed over 50% reduction compared to the precursor cells (Fig. 2D). For macrophages, η₁ = 174.0±10.2 Pa s, while neutrophils and monocytes had η₁ = 212.5±43.3 Pa s and 142.1±44.7 Pa s, respectively. It is intriguing that we cannot obtain η₁ for undifferentiated cells within the 4 s of stretching but can do so for the differentiated cells. Extrapolating the viscoelastic model up to 20 s predicts a steady-state viscosity of over 300 Pa s for the undifferentiated cells. To obtain their steady-state viscosity directly within the actual data range, we carried out 20 s creep tests for the undifferentiated cells (N = 3 experiments) and found η₁ = 333.5±26 Pa s, which is more than 50% higher than the values for all three differentiated cell types. The lower η₁ of all differentiated cells implies that beyond some transient regime, all three differentiated cells should be more compliant than undifferentiated cells. We therefore validated this implication by also performing 20 s creep tests for macrophages, which are less compliant than undifferentiated cells at the 4 s creep testing. Indeed, the macrophages are significantly more compliant at 20 s creep testing (Fig. 2E). Succinctly, monocytes and neutrophils are elastically compliant cells at short time-scales. The lower steady-state viscosity is an apparent indication of their ‘latent’ fluid-like behavior at long time-scales. Macrophages, on the other hand, are generally more fluid-like cells and are less compliant than undifferentiated cells at short time-scales but more compliant at long time-scales.

Material properties of cells are function-dependent

These distinct rheological properties, robustly emerging towards the respective end of the differentiation into the mature phenotypes, suggest an intriguing hypothesis, where the material properties might be physical prerequisites for specific functions of the mature cell types. After all, circulating neutrophils and monocytes have to repeatedly squeeze through microvasculature constrictions smaller than their diameter while being advected along at high speed [19] (Fig. 3Aa). Since these are rapid impact processes, the cells need to be elastically compliant at short time-scales (<1 s), which the macrophages are not. To test this prediction we used pressures in the physiological range (5–20 mbar) to send cells through channels in PDMS with inner cross-sections of 10 µm×12 µm and 12 µm×12 µm, fabricated by photolithography [20] using PDMS (see Methods S1). To pass through the 12 µm×12 µm channel at a pressure of 20 mbar, macrophages required an average advection time (from entry to exit) of 3.98±1.77 s, an order of magnitude longer than all other cell types (Fig. 3B; p<0.05). At this pressure, macrophages could hardly be advected at all through the narrower 10 µm×12 µm channel (see Video S1) although all other cell types went through (see Video S2) even at a lower pressure of 10 mbar (Fig. S5). However, macrophages treated with 2 µM cytochalasin D, which depolymerizes filamentous actin [21], were 40% more compliant (Fig. 4A) and had an average advection time of 0.34±0.11 s, which is not significantly different from those of neutrophils, monocytes and precursors (Fig. 3B). These results confirm that cell compliance is the major limiting factor controlling the rapid passage of cells through narrow constrictions at short time-scales.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Time-scales and niche-specific relevance of cell viscoelasticity.

A, In vivo, neutrophils and monocytes navigate constrictions smaller than themselves (y) and deform rapidly as illustrated (<1 s). Macrophages on the other hand, as well as neutrophils and monocytes after extravasation, migrate through tissue, which occurs over much longer time-scales. B, Short time scale advection of cells through the 12 µm×12 µm channel at a pressure of 20 mbar. Macrophages (n = 50) require an average advection time (from entry to exit) of 3.98±1.77 s, an order of magnitude longer than all other cell types and with statistically significant difference (p<0.05). There are no statistically significant differences between the advection times of HL60 (n = 24), neutrophils (n = 36), monocytes (n = 50) and macrophages treated with 2 µM Cytochalasin D shown as Mac+ (n = 67). Scale bar is 10 µm. C, Long time scale migration of the three mature cell types and precursors in Boyden chamber assays, where cells squeeze actively through a thick, porous membrane. After 3 h, significantly (p<0.0001) more of all three differentiated cell types have migrated than the undifferentiated, more viscous cells. Scale bar is 10 µm.

https://doi.org/10.1371/journal.pone.0045237.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. Pharmacological and functional assays and primary cell results.

A, The cytoskeletal drug cytochalasin D (cyto D) disrupts filamentous actin, leading to increase in compliance (over 40%) due to actin depolymerization in all cell types. B, As a functional assay using the chemoattractant fMLP, all differentiated cell types responded to fMLP in a manner consistent with their in vivo counterparts. C, Primary human neutrophils and CD34+-derived neutrophils are respectively about 3-fold and 4-fold more compliant than CD34+ primary human cord blood stem cells. The **** indicates a significant difference with a p value of <0.0001.

https://doi.org/10.1371/journal.pone.0045237.g004

As phagocytes, all three mature cell types respond to microbial infection and so have to migrate to and through tissues to sites of infection. The crowded environment of tissues, as well as the absence of fluid pressure as the driving force (as in blood), makes tissue migration a process occurring at much longer time-scales (Fig. 3A). To test whether the cells' material properties are also in line with this second cell function, we compared the migratory ability of the three mature cell types with that of their precursors in Boyden chamber assays [5], where cells have to squeeze actively through a thick, porous membrane (Fig. 3C and Methods S1). After 3 h, more cells of all three differentiated lineages had migrated through the pores than the undifferentiated cells. Of interest, macrophages and monocytes (which ultimately become macrophages in vivo) both showed a higher relative increase in migration ability of 204% and 194%, respectively, compared to 145% for neutrophils (Fig. 3C). It is quite remarkable that an in vitro migration assay, which mimics in vivo migration in tissue over long times, correlates with the lowered steady-state viscosity of differentiated myeloid cells. The obvious interpretation is that the lowered steady-state viscosity is a physical prerequisite for ‘flowing’ through tissues over long time-scales (Fig. 3A). The fact that macrophages do not retain the high compliance of monocytes at short time-scales suggests an adaptation of their mechanical property to their cellular niche: they are tissue-resident cells and squeezing through tissue is a long-time process which does not require short-time compliance. These in vivo exigencies appear to explain our results, engendering insights into cell circulation and migration behaviour. In fact, recent in vitro studies show that matrix structural configuration determines 3D migration modes of human macrophages [22]. Our results therefore suggest novel cell mechanical-property-based therapeutic targets for blood/immune system dysfunctions where aberrant circulatory and migratory functions of neutrophils and monocytes/macrophages are implicated, such as sepsis [23], artherosclerosis [15] and chronic obstructive pulmonary disease [24].

Confirmation of results with functional assays and primary cells

To ascertain the biological relevance of our results we tested whether the compliance measured is dictated by the cytoskeleton of cells, by disrupting filamentous actin using cytochalasin D. F-actin is a major determinant of cellular mechanical properties in vitro and in vivo [25]. Indeed, in all cell types we found over 40% increase in compliance due to actin depolymerization (Fig. 4A). Also the transition time dropped from 5 s or 15 s to less than 0.8 s in all cell types (Fig. S6).

Furthermore, we carried out myeloid functional assays on the differentiated cells using the chemo-attractant N-formyl-methionine-leucine-phenylalanine (fMLP). In vivo monocytes, neutrophils and macrophages express membrane receptors for bacterial fMLP which allow them to respond to chemotactic gradients, while fMLP-mediated chemotaxis is not detectable with undifferentiated HL60 cells [26]. Indeed, all three differentiated cells types (but not the undifferentiated cells) showed altered mechanical properties in response to in vitro fMLP treatment (Fig. 4B).

Since cell lines are only models of primary cells, we also measured and compared the compliance of primary human hematopoietic stem cells (CD34+) with in vitro differentiated CD34+ cells as well as with primary human neutrophils and monocytes. Consistent with the cell line results, the primary human neutrophils and the CD34+-derived neutrophils were respectively 3-fold and 4-fold more compliant than primary CD34+ stem cells (Fig. 4C). The primary monocytes were also significantly more compliant than primary stem cells (Fig. S7). Furthermore, macrophages derived from the differentiation of CD34+ stem cells were significantly (p<0.01) less compliant than undifferentiated CD34+ stem cells (Fig. S8). Of note, since monocytes turn into macrophages in vivo, we also measured primary monocytes before and after they had turned into macrophages in culture and found the latter to be 40% less compliant than the monocytes, consistent with all other results reported above. Thus, cell compliance is a lineage-specific characteristic material property of both cell lines and primary myeloid cells.

Ethics Statement

Ethical approval was obtained from the Cambridgeshire 2 Research Ethics Committee, and written informed consent, or written assent from next-of-kin where appropriate, were obtained in all cases in accordance with the Declaration of Helsinki.

Cell culture and differentiation

We used HL-60/S4 cells [34], a rapidly differentiating clone or sub-line of HL-60 cells, as a model for myeloid differentiation. The cell line was a gift from Donald and Ada Olins of the Department of Pharmaceutical Sciences, College of Pharmacy, University of New England. Cells were cultured and induced to differentiate along the myeloid lineage following well-established protocols [33]. Differentiation was confirmed in various ways (see Methods S1 and Fig. S1). Primary human neutrophils and monocytes were isolated from peripheral blood taken from healthy donors at the Department of Medicine, Addenbrooke's Hospital, University of Cambridge. Human cord blood-derived CD34+ stem cells were obtained from Translational Research Laboratory, Department of Haematology, University of Cambridge. Details of cell culture and differentiation for both the cell line model and primary HSCs as well as the preparation of primary neutrophils and monocytes can be found in Methods S1.

Optical stretching and determination of creep compliance

The principle, setup and handling of the optical stretcher have been described previously [5], [16], [35]. The forces that trap and deform the cell outwardly along the surface arise from the change in the refractive index at the cell-medium interface and the ensuing conservation of momentum [36]. The axial strain during optical stretching is given by(4)where a_o is the semi-major axis of the unstretched cell and a(t) is the time-varying semi-major axis measured (Fig. 1B). The optical stress on the cells is computed using an electromagnetic wave model [37], which requires knowledge of the cells' average refractive index. We measured the refractive index of the cells during differentiation using a digital holographic microscope [38]. The strain is normalized by the peak value of the calculated optical stress σ₀ and a geometric factor F_g to give the creep compliance.(5)The geometric factor is calculated as described elsewhere [18] to account for cell size and stress distribution. Of note, cell size changes characteristically upon differentiation (see Fig. S3). Thus, by this normalization the mechanical properties reported here are decoupled from the variation in the dielectric characteristics and in the size of cells during differentiation.

Statistical analysis

For each cell type or treatment condition the number of cells per OS experiment was n≥30. Compliance data are presented as mean ± SEM. Representative compliance data were chosen from total number of independent experiments, N>5 for all cell lines and N≥3 for primary cells. Overall, the number of individual cells analyzed totaled about 5000. The rheological parameters J_o and β were confirmed in two independent experiments. For each cell type or treatment condition the number of cells per advection experiment was n>20. Advection times are representative of 3 independent experiments and are presented as mean ± SEM. Statistical comparisons for all experiments were done using one-way ANOVA in Origin (OriginLab Corporation, Northampton, MA, USA), to account for multiple groups. Within ANOVA, significant differences were reported only where at least three different means comparison tests (Tukey, Bonferroni and Dunn-Sidak) simultaneously showed such differences. All box plots include whiskers at 5–95% range, horizontal box lines at 25–75% range, median line and the mean (inset box). In the parametric fitting 95% confidence intervals were computed, and extracted parameters were considered significant where confidence intervals did not include 0. All curve fitting was performed using Matlab (Mathworks, Natwick, MA).