Tuning intercellular adhesion with membrane-anchored oligonucleotides

Abstract

Adhesive interactions between cells play an integral role in development, differentiation and regeneration. Existing methods for controlling cell–cell cohesion and adhesion by manipulating protein expression are constrained by biological interdependencies, e.g. coupling of cadherins to actomyosin force-feedback mechanisms. We use oligonucleotides conjugated to PEGylated lipid anchors (ssDNAPEGDPPE) to introduce artificial cell–cell adhesion that is largely decoupled from the internal cytoskeleton. We describe cell–cell doublets with a mechanical model based on isotropic, elastic deformation of spheres to estimate the adhesion at the cell–cell interface. Physical manipulation of adhesion by modulating the PEG-lipid to ssDNAPEGDPPE ratio, and conversely treating with actin-depolymerizing cytochalasin D, resulted in decreases and increases in doublet contact area, respectively. Our data are relevant to the ongoing discussion over mechanisms of tissue surface tension and in agreement with models based on opposing cortical and cohesive forces. PEG-lipid modulation of doublet geometries resulted in a well-defined curve indicating continuity, enabling prescriptive calibration for controlling doublet geometry. Our study demonstrates tuning of basic doublet adhesion, laying the foundation for more complex multicellular adhesion control independent of protein expression.

1. Introduction

Heterotypic cell–cell adhesion and homotypic cohesion are increasingly recognized as important practical control parameters in tissue engineering applications [1–3] and in guidance of cell behaviours such as differentiation [4–6]. However, cell–cell adhesions mediated by membrane proteins (cadherins, integrins and the immunoglobulin superfamily) are inextricably connected to other biochemical processes. Cadherins interface with cytoskeletal actin on the cytosolic side of the membrane associating with force-generating processes such as migration, mechanosensing and ECM adhesion [7–10]. Artificial control of protein-mediated adhesion can cause unintended consequences in coupled processes such as differentiation, proliferation and apoptosis [11–13], demanding that one negotiates such interdependency by fixing variables to reduce degrees of freedom [14–16].

Manipulating expression of interacting proteins remains the dominant strategy for modulating cohesive interactions between cells [17–19], though a converse strategy is to restrict the spatial positioning of cells with 2D and 3D localization methods such as micropatterning to induce natural cohesion in a predictable manner [20–24].

Cell surface modification with phospholipid-anchored oligonucleotides (ssDNAPEGDPPE) offers a means of inducing physical attachment of cells to substrates or other cells orthogonal to natural adhesion biochemistry [25–29]. Recently, Teramura [30] showed that ssDNAPEGDPPE could be used to investigate the interactions of adhering cell doublets demonstrating dependence of cell–cell interactions on their degree of contact. We have also previously shown that native integrin-mediated adhesion modes take place in parallel with artificially induced attachment with ssDNAPEGDPPE [28].

The goal of this study is to use artificial ssDNAPEGDPPE-mediated adhesion in a cell doublet model as a tool to manipulate geometric parameters, e.g. their interfacial contact area and extent of membrane deformation, while minimizing the effects of coupling to mechanical, cytoskeletal biochemistry characteristics of other adhesion modes. We show that by tuning the concentration of ssDNAPEGDPPE molecules on the surface of cells before doublet formation, it is possible to achieve a well-defined range of doublet geometry distributions characterized by the interfacial contact area.

2. Material and methods

2.1. Supplies and reagents

α-N-hydroxysuccinimidyl-ω-maleimidyl poly(ethylene glycol) (NHS-PEG-Mal, MW 5000), α-succinimidyl carbonyl-ω-methoxy, polyoxyethylene (meoPEGDPPE) and 1,2-dipalmitoyl-sn-glycerol-3-phosphatidylethanolamine (DPPE) were purchased from NOF Corporation (Tokyo, Japan). Fetal bovine serum, Hanks' balanced salt solution (HBSS) and RPMI 1640 medium were purchased from Invitrogen, Co. (Carlsbad, CA, USA). Cytochalasin D was purchased from Sigma Aldrich (St Louis, MO, USA).

Single-stranded DNA (ssDNA) with 5′ terminal disulfide modifications and ssDNA with 5′ disulfide and 3′ fluorescein amidite were purchased from Fasmac (Atsugi City, Kanagawa Japan).

2.2. Modification of cells with ssDNAPEGDPPE

The cells used in this experiment were CCRF-CEM cells, a human T-cell lymphoblast-like cell line and non-adherent cell type were used in previous ssDNAPEGDPPE experiments [31]. Oligo-terminated phospholipids (ssDNAPEGDPPE) (figure 1a) [31,32] were synthesized in two steps as previously described by first conjugating NHS-PEG-maleimide to DPPE followed by a second conjugation of maleimide-PEGDPPE to thiol-terminated oligonucleotides. PEGDPPE was produced with an analogous single step conjugation of PEGNHS to DPPE. Cells were modified with ssDNAPEGDPPE by incubation for 1 h, RT, in a solution of ssDNAPEGDPPE at a concentration of 200 μg ml⁻¹. Cells were rinsed thrice with HBSS followed by pelleting. Doublets were formed by mixing complementary ssDNAPEGDPPE-bearing cell groups into a single glass bottom dish during observation with confocal microscopy. Oligonucleotides anchored to cells were visualized by substituting SeqAPEGDPPE for fluorescein-terminated FAMSeqAPEGDPPE (table 1).

Figure 1.

(a) Chemical structure of ssDNAPEGDPPE. (b) Tomographic reconstruction of fluorescently labelled cell doublet co-adhering through FAM-modified ssDNAPEGDPPE hybridization across adjacent membranes. (c) Coarse-grained geometric representation of an ideal cell doublet: 2 intersecting spheres with truncated regions corresponding to deformed elastic cells. (d) Surface plot of potential energy as a function of doublet interfacial contact area and interfacial adhesion and a colour map representative of the force along the centre-to-centre axis. (Online version in colour.)

Table 1.

Thiol-modified ssDNA Sequences.

sequence name	sequence 5′ to 3′
SeqA	SH-TGCGGAGGATCCTTTCACACA
SeqA-FAM	SH-TGCGGAGGATCCTTTCACACA-FAM
SeqA’	SH-TGTGTGAAAGGATCCTCCGCA

2.3. Formation of cell doublets

Cell doublets were formed after preparing two parallel populations of modified cells bearing (FAM)-SeqAPEGDPPE and its complement SeqA’PEGDPPE such that modified cells could form complementary pairings upon mixing of the two populations. Complementary cell groups were exposed to each other over a glass bottom dish during imaging by a confocal laser scanning microscopy and allowed to settle after aggregation to reach steady state over the course of 15 min. Doublets were identified manually out of a population also containing single cells, triplets and larger aggregates. A tomographically reconstructed 3D image of a ssDNAPEGDPPE-co-adhered cell doublet with fluorescently labelled ssDNA is shown in figure 1b.

For cells modified to exhibit reduced adhesion, methoxy-PEGDPPE (meoPEGDPPE) was added to the ssDNAPEGDPPE solutions during the cell modification stage in various molar ratios to provide binding competition and reduce the equilibrium ssDNAPEGDPPE membrane surface concentration. Co-PEGDPPE/ssDNAPEGDPPE-modified cells were combined in complementary pairs as with ssDNAPEGDPPE-modified cells to form a distribution composed of single cells, doublets, triplets and high-order cell–cell aggregates from which doublets were manually identified and imaged with confocal microscopy.

For cells treated pharmacologically to exhibit reduced elastic repulsion, cytochalasin D (1 mg ml⁻¹ in DMSO) was added to cell suspensions during the 1 h ssDNAPEGDPPE modification for a final concentration of 2 μg ml⁻¹ [33], and cells observed with confocal microscopy were kept in a medium containing 2 μg ml⁻¹ cytochalasin D.

2.4. Estimation of adhesion via doublet geometry

Cells i and j interact with a total energy

$$W=W_{\mathrm{Rep}}+W_{\mathrm{Adh}},$$ 2.1

where W_Rep is the potential energy contribution due to elastic repulsion and W_Adh is the potential energy contribution resulting from adhesion over the surface of contact.

To estimate the adhesion energy, we have chosen to model the cell-repulsive energy due to elastic deformation by the Hertz model applied to two compressed elastic spheres described in Landau and Lifschitz’s Theory of elasticity [34,35] shown schematically in figure 1c. Two cells i and j of unequal spherical radii R_i and R_j are divided by an interfacial contact area A_contact of diameter a_ij. The two cell centres of curvature are separated by a distance d_ij, and the truncated caps due to adhesion are together of length x_ij referred to as the indentation depth. The contribution to interfacial energy due to repulsion is

$$W_{\mathrm{Rep}}=\frac{2}{15}{(R_i+R_j-d_{ij})}^{5/2}{\left(\frac{1-{\nu }^2}{\tilde{E}}\right)}^{-1}\sqrt{\frac{R_i{R}_j}{R_i+R_j}},$$ 2.2

where $\tilde{E}$ and ν are the Young’s modulus and Poisson’s ratio, respectively, defined for both cells i and j.

We assume that the radii of curvature of cells can be represented by spherical radii. We imported the Young’s modulus, which we assume for simplicity to be constant in the Hertz model, from Rosenbluth et al. (E = 855 ± 670 Pa for HL60 leukaemia myeloid cells), and likewise assumed a Poisson’s ratio of 0.5 as was done by Rosenbluth et al. [36].

The adhesive contribution to potential energy is

$$W_{\mathrm{Adh}}=\gamma A_{\mathrm{contact}},$$ 2.3

where γ is the adhesion represented in units of energy per unit area. The contact area can be rewritten as a function of R_i, R_j and d_ij such that the potential due to adhesion becomes

$$W_{\mathrm{Adh}}=\gamma \pi (R_i^2-\frac{{(d_{ij}^2-R_j^2+R_i^2)}^2}{4d_{ij}^2}).$$ 2.4

Thus, the given values of R_i, R_j and γ, the total interfacial energy between cells i and j can be determined as a function of d_ij or x_ij. Figure 1d shows a plot of potential energies W (colour map), W_Rep (red) and W_Adh (blue) as a function of interfacial contact area A_contact and adhesion γ for a hypothetical pair of fixed volumes V_i = V_j = 900 μm³.

The minimum of W can be determined by solving for the root of its derivative with respect to separation distance, the point where the force along the axis extending from cell centre to centre is zero

$$\begin{array}{rl}\frac{\text{d}W}{\text{d}{d}_{ij}}& =\frac{-\gamma \pi }{2d_{ij}^3}(d_{ij}^4-{(R_i^2-R_j^2)}^2)\\ & -\frac{1}{3}{\left(\frac{1-{\nu }^2}{\tilde{E}}\right)}^{-1}\sqrt{\frac{R_i{R}_j}{R_i+R_j}}{(R_i+R_j-d_{ij})}^{3/2},\end{array}$$ 2.5

and is shown in figure 1d as the colour map. For values of γ sufficient to overcome the thermal and convective noise, a potential energy well exists for W such that A_contact > 0 and cell adhesion occurs at steady state. This region is indicated by white on the colour map in figure 1d. The roots of the derivative equation can be rearranged to solve for adhesion

$$\begin{array}{rl}\gamma & =-\frac{2}{3\pi }\sqrt{\frac{R_i{R}_j}{R_i+R_j}}{\left(\frac{1-{\nu }^2}{\tilde{E}}\right)}^{-1}\hspace{0.17em}\hspace{0.17em}\frac{d_{ij}^3{(R_i+R_j-d_{ij})}^{3/2}}{d_{ij}^4-{(R_i^2-R_j^2)}^2},\\ \frac{\text{d}W}{\text{d}{d}_{ij}}& =0.\end{array}$$ 2.6

To compute the space of possible steady state doublet conformations including contact area as a function of both the Young’s modulus and adhesion (figure 2d), a numerical algorithm was implemented to identify the values of A_contact which satisfy the root of (dW/dd_ij) under an additional constraint of fixed cell volume (electronic supplementary material, figure S1).

Figure 2.

Area of the interfacial contact plane between cell doublets (A_contact) can be influenced by modulating parameters: Young’s modulus or adhesion. (a) Cross-section of doublet formed from cells modified with a mixture of meoPEGDPPE and ssDNAPEGDPPE resulting in a greatly reduced ${\overline{A}}_{\mathrm{contact}}$. (b) Normal doublet formed form cells modified with only DNAPEGDPPE. (c) Doublet with large ${\overline{A}}_{\mathrm{contact}}$ formed from DNAPEGDPPE-modified cells treated with cytochalasin D. Scale bar = 10 μm. (d) Surface plot of ${\overline{A}}_{\mathrm{contact}}$ (vertical axis) as a function of the Young’s modulus ($\tilde{E}$) and the adhesion (γ) with the surface energy W indicated by the colour map. Arrows indicate hypothetical changes in ${\overline{A}}_{\mathrm{contact}}$ reached by orthogonal movements along the parameter axes. (e) Histogram of ${\overline{A}}_{\mathrm{contact}}$ measurements determined from cross-sectional images of doublets formed under normal conditions, with the addition of meoPEGDPPE at various relative concentrations (meoPEGDPPE : ssDNAPEGDPPE ratios) during cell modification, and with the addition of cytochalasin D (blue). (f) Histogram of adhesion values calculated from the geometry of doublet cross-sections based on a fixed and constant $\tilde{E}$ for doublets formed under normal conditions and doublets with various relative concentrations of meoPEGDPPE. (meoPEGDPPE : ssDNAPEGDPPE ratios). (g) Plot of the absolute value of adhesion $|\gamma |$ calculated on the basis of doublet geometry versus the molar fraction of meoPEGDPPE to ssDNAPEGDPPE used during cell modifications. (Online version in colour.)

2.5. Doublet image processing

A partially automated image processing algorithm was applied to doublet images in order to extract their geometric parameters and the fluorescence intensity due to FAM-SeqAPEGDPPE at interfaces (electronic supplementary material, figure S2). Cropped interface images (figure 4a) were then processed with a numerical signal detection algorithm (see electronic supplementary material), which begins by shifting the alignment of rows of pixels along the vertical axis of the image such that the brightest pixel in each row is centred (figure 4b). This action corrects for irregularities in shape of the interface leading to a sharper distribution of pixel intensities when the image is collapsed into a single horizontal axis. The collapsed intensity profile peak is then integrated and divided by the interface diameter or chord length a_ij to obtain the intensity density of the profile in arbitrary intensity units.

Figure 4.

Image analysis of doublet interface cross-sections. (a) Raw image of a doublet interface cross-section taken with confocal microscopy. Scale bar: 10 μm. (b) Summed pixel intensities from aligned profile (inset) collapsed onto a single x-axis. (c) Contact area measurements for cytochalasin D-treated doublets plotted against their computed intensity density values at different meoPEGDPPE : ssDNAPEGDPPE molar ratios. (d) Absolute value of adhesion population means and steady state contact area population means plotted versus their corresponding mean intensity densities. (e) Intensity density population means of image-processed doublet interfaces (red) and single cell membranes (blue) plotted as a function of the molar ratio of meoPEGDPPE to ssDNAPEGDPPE. (Online version in colour.)

2.6. Fluorescence spectroscopy of digested FAM-ssDNAPEGDPPE

UV/Vis spectroscopy was carried out using a Beckman DU-640 spectrophotometer (Beckman Coulter, Brea, CA, USA) and fluorescence spectroscopy with a Hitachi F-2710 fluorometer (Hitachi, Tokyo, Japan). To calibrate the concentration estimates of the prepared solutions of FAM-ssDNAPEGDPPE, a standard was constructed with UV/Vis absorbance measurements taken at 260 nm and adjusted to account for the absorbance contribution from FAM. A linear calibration curve (electronic supplementary material, figure S3a) was used to adjust concentration estimates used in later steps. A standard was then constructed with fluorescence spectroscopic measurements at several different concentrations of FAM-SeqAPEGDPPE digested with Benzonase [25] (1x RQ1 buffer (Promega, Madison, WI, USA) supplemented 0.9% NaCl and 2500 U ml⁻¹ Benzonase, and incubated for 10 min at 37°C) to mimic the conditions of cell experiments requiring the liberation of FAM from FAM-SeqAPEGDPPE-modified cell membranes (electronic supplementary material, figure S3b). The coefficients of the linear standard, k_spec = 3.71 ± 0.04 × 10⁻⁴ for the slope and β_spec = 9.59 ± 1.8 × 10⁻² for the intercept, were then used to infer the FAM concentrations of supernatants extracted from suspensions of cells bearing FAM-SeqAPEGDPPE following DNA digestion with Benzonase.

Suspensions of FAM-SeqAPEGDPPE-bearing cells were first prepared at concentrations of approximately 200 cells μl⁻¹, washed, pelleted and resuspended in 1x RQ1 buffer (Promega, Madison, WI, USA) supplemented 0.9% NaCl and 2500 U ml⁻¹ Benzonase and incubated for 10 min at 37°C. The cells were again pelleted and the supernatant containing liberated FAM and digested nucleotide fragments was extracted for spectroscopic analysis. Prior to digestion, a fraction of suspensions were analysed with flow cytometry to estimate the total number of cells in the Benzonase-treated samples. Supernatant fluorescence spectroscopy was performed using 495 nm excitation with a detection range spanning 510–550 nm.

2.7. Flow cytometry analysis of FAM-ssDNAPEGDPPE-modified cells

Cells sampled from the same non-aggregated populations of FAM-SeqAPEGDPPE-modified cells, as those analysed by a confocal microscopy and a post-digestion fluorescence spectroscopy, were sampled with a flow cytometer (5000 cells, Guava Easy Cyte Mini, Millipore, Billerica, MA, USA). Flow cytometry-based measurement of number of cells per microlitre of suspension were used to determine the number of cells treated with Benzonase for fluorescence spectroscopy analysis of liberated FAM.

2.8. Single cell imaging and membrane fluorescence profile quantification

Confocal microscopic profiles of non-aggregated single cells modified with FAM-SeqAPEGDPPE were analysed in a semi-automated fashion. Manually identified centres of the cells in cross-section images were used as the origin of a polar to Cartesian transformation of the fluorescent halos visible along cell membranes with intensities, indicating relative amounts of FAM-SeqAPEGDPPE on the cell surface. Fluorescence profiles were aligned on a single vertical axis to increase the sharpness of profile peaks prior to the integration of the peak for single quantification of the fluorescence intensities (see electronic supplementary material).

2.9. Imaging conditions

Mixed complementary groups of FAM-SeqAPEGDPPE-modified and SeqA’PEGDPPE-modified cells were briefly agitated to induce aggregation by hybridization and pipetted onto glass bottom dishes and allowed to settle and reach a steady state for 15 min prior to imaging with a Fluoview FV500 confocal laser scanning microscope. Steady state was reached on the microscope sample stage to prevent further contact between unhybridized cells and cell aggregates. Doublets were identified manually and imaged with a 60 × 1.35 NA oil objective. The z-depth of each cross-sectional image was adjusted manually to obtain an image of the maximally long central chord of the doublet interfacial contact area. While this procedure ensures unbiased measurement of the interfacial contact diameter a_ij, it does not account for slanting of the axis between the two cells that could be caused by a difference in cell radii. An adjustment, based on geometric constraints, is applied then to values of d_ij, R_i and R_j obtained experimentally (see electronic supplementary material). Confocal microscopic profiles of non-aggregated single cells modified with FAM-SeqAPEGDPPE were imaged by manually scanning z-depth to locate the profile with maximal radius and imaging plane intersecting with the cell centre.

2.10. Statistics

Error propagation, according to the variance equation, was estimated for molecular densities, adhesion measured via geometry and steady state interfacial contact area values. Error bars for these computed quantities represent propagated errors, and error bars in all other cases indicate standard deviations.

3. Results

3.1. Development of coarse-grained cell doublet model

By modifying complementary batches of cells respectively with oligonucleotide-conjugated PEG-lipids (SeqAPEGDPPE and SeqA’PEGDPPE; figure 1a) followed by suspending the two populations together, we formed doublets along with other aggregates of CCRF-CEM cells co-adhered at an adjoining interface by the hybridization of complementary single-stranded oligonucleotides anchored to the cell membranes. By substituting SeqAPEGDPPE with a fluorescently labelled analogue (FAM-SeqAPEGDPPE), it was possible to visualize the distribution of ssDNAPEGDPPE of one of the cells in each doublet and construct a tomographic picture from multiple confocal laser scanning microscopic sections (figure 1b). Based on this information and the precedent set by similar models of cadherin-mediated cohesion [37–40], we have developed a coarse-grained model of the DNAPEGDPPE-co-adhered cell doublet consisting of two intersecting spheres of unequal radii and plane of intersection, representing the interface of adhesion between two deformed elastic spheres (figure 1c). The dynamics of the doublet is driven by the imbalance of repulsive elastic force and attractive adhesive force in the direction of the centre-to-centre axis (equation (2.5)), and when interfacial adhesion is greater than zero, a non-zero interfacial contact area exists corresponding with the potential energy minimum where repulsive and attractive forces are balanced (figure 1d).

We designed and implemented a numerical algorithm to compute the interfacial contact area A_contact at steady state (${\overline{A}}_{\mathrm{contact}}$), a key geometric parameter indicative of the extent of cohesive deformation for a given cell type of known size distribution. The interfacial contact area A_contact is at steady state when attractive and repulsive forces are balanced for two cells of fixed volume co-adhered according to our doublet model as a function of the Young’s modulus $\tilde{E}$ and the adhesion parameter γ or adhesion energy per unit area at the interface. Examining the space of ${\overline{A}}_{\mathrm{contact}}$, we can see that doublets formed with low γ and large $\tilde{E}$ are smallest corresponding to weakly adhesive stiff cells. Conversely, doublets with large γ and low $\tilde{E}$, in other words, soft and highly adhesive cells, have the largest predicted ${\overline{A}}_{\mathrm{contact}}$ and deepest potential energy well. To test our model, we wished to explore this space of ${\overline{A}}_{\mathrm{contact}}$’s by modulating γ and $\tilde{E}$ independently of each other.

3.2. Experimental investigation of independent elastic and cohesive parameters of doublet geometry

We prepared cell doublets with reduced interfacial adhesion by reducing the surface concentration of ssDNAPEGDPPE by the introduction of non-hybridizing methoxy-PEG-lipids during the cell modification stage to compete with ssDNAPEGDPPE for binding sites on the cell membrane. When cells were modified in buffer containing a 10 : 1 molar ratio of meoPEGDPPE to FAM-SeqAPEGDPPE/SeqA’PEGDPPE (figure 2a), confocal images of resulting doublets revealed cross-sections of ${\overline{A}}_{\mathrm{contact}}$ values qualitatively smaller than those of normal doublets prepared from cell groups modified without meoPEGDPPE (figure 2b).

We wished to reduce the mechanical integrity of cells, so we treated cells during the ssDNAPEGDPPE modification stage and during doublet formation/imaging with cytochalasin D, a cytoskeletal inhibitor previously shown prevent actin polymerization and reduce cortical integrity [41,42]. Doublets prepared with cytochalasin D (figure 2c) exhibited qualitatively larger ${\overline{A}}_{\mathrm{contact}}$ cross-sections in confocal images compared to normal doublets (figure 2b) and a similar behaviour was noted for triplets and higher-order structures observed (electronic supplementary material, figure S4). This is in line with the landscape of our model parameter space (figure 2d).

In addition to the 10 : 1 meoPEGDPPE : ssDNAPEGDPPE, normal (0 : 1 meoPEGDPPE : ssDNAPEGDPPE) and cytochalasin D-treated preparations, we prepared doublets from cells modified with 5 : 1, 2 : 1 and 1 : 1 meoPEGDPPE : ssDNAPEGDPPE molar ratios and analysed confocal images of n = 50 doublets per condition by manually tracing circles along the contours of doublet cross-sections from which the geometric parameters d_ij, R_i and R_j could be computed. From these values, interfacial diameters a_ij and ${\overline{A}}_{\mathrm{contact}}$ of each doublet were computed and compiled into the histogram shown in figure 2e. Provided our assumed values for the Young’s modulus and Poisson’s ratio, we computed the adhesion from the assumption that cohesive and elastic forces are balanced (equation (2.6)) and compiled these into a histogram in figure 2f . Since we do not know $\tilde{E}$ of the cytochalasin D-treated doublets relative to our postulated $\tilde{E}$, this condition was excluded from adhesion computation. Despite large standard deviations in ensemble adhesion estimates ($\text{SE}>60\text{\%}$), there was a strong correlation between the molar ratio of meoPEGDPPE : ssDNAPEGDPPE and the adhesion computed on the basis of our geometric model (linear fit: R² = 0.99). Large sample sizes permit prescriptive control over the average doublet geometry (SEM < 10%) (figure 2g).

3.3. Fluorescence-based quantification of hybridizable molecular surface concentration

In our mechanical model of the cell doublet, we infer the adhesion at the cell–cell interface on the basis of force balance with elastic forces in the absence of knowledge of the molecular distribution. To better understand the distribution of ssDNAPEGDPPE molecules at the cell–cell interface and their contribution to adhesion, we use the fluorescence of fluorescein-terminated FAM-SeqAPEGDPPE to track the distribution and concentration of ssDNAPEGDPPE molecules in the cell membrane. We first examined single cells prepared with mixtures of meoPEGDPPE (meoPEGDPPE; figure 3a) and FAM-SeqAPEGDPPE with flow cytometry (figure 3b). Forward scatter versus green fluorescence distributions were of highly reproducible shape between conditions besides being shifted in their peak fluorescence according to the meoPEGDPPE : FAM–SeqAPEGDPPE ratio used in their preparation.

Figure 3.

Modulation of meoPEGDPPE to ssDNAPEGDPPE ratio can be used to tune the density of hybridizable molecules on cell surfaces. (a) Structure of meoPEGDPPE. (b) Flow cytometry results displayed as forward scatter values versus green fluorescence. (c) Schematic illustration of a cell saturated with PEG-lipids, some of which terminate as methoxy-PEG, while others are connected to FAM-labelled ssDNA. (d) Schematic illustration of FAM extraction by digesting nucleotides with Benzonase followed by cell separation from FAM-containing supernatant by centrifugation. (e) Before the image of FAM-SeqAPEGDPPE-modified cells. (f) After the image of FAM-SeqAPEGDPPE-modified cells following digestion with Benzonase. Scale bar = 10 μm. (g) Plot mean of flow cytometry fluorescence measurements of cell populations modified with different relative molar fractions of meoPEGDPPE to FAM-SeqAPEGDPPE. Linear fit R² = 0.99. (h) Plot of population mean molecular surface concentrations computed on the basis of fluorescence spectroscopy of supernatants versus relative molar fraction of meoPEGDPPE to ssDNAPEGDPPE. Linear fit R² = 0.93. (i) Plot population mean of flow cytometry fluorescence measurements versus population mean fluorescence intensity densities computed from confocal image analysis. Linear fit R² = 0.96. (j) Plot of population mean molecular surface concentrations computed on the basis of fluorescence spectroscopy of supernatants versus population mean fluorescence intensity densities computed from confocal image analysis. Linear fit R² = 0.88. (Online version in colour.)

We developed a fluorescence spectroscopy-based assay to estimate the surface concentration of FAM-SeqAPEGDPPE on membranes of single cells sampled from the same population of cells used for flow cytometry. We used oligonucleotides terminated with a single fluorescein moiety to ensure that the fluorescent emission corresponds to oligonucleotide (SeqA) concentration while leaving meoPEGDPPE molecules unlabelled (figure 3c). We used Benzonase, a non-specific nuclease capable of digesting ssDNA and dsDNA, to digest away the oligonucleotide-tethered fluorescein molecules, releasing them into the surrounding medium which we collected and analysed with fluorescence spectroscopy (figure 3d). Fluorescent halos marking the contour of the cell membrane in confocal cross-sections could be observed after cells were modified with FAM-SeqAPEGDPPE (figure 3e), while after Benzonase digestion and removal of the digest supernatant the remaining cells exhibited negligible fluorescence levels when re-imaged for qualitative characterization (figure 3f ) indicating that the fluorescein was extracted. Based on a linear calibration relating solution fluorescence intensity to the solution ssDNA concentration (see electronic supplementary material), we measured the total amount of fluorescein extracted and could then estimate the average FAM-SeqAPEGDPPE concentration per cell by dividing by the number of cells used in the extract preparation. The molar ratio of meoPEGDPPE to SeqAPEGDPPE correlates inversely with the mean peak fluorescence of cell populations analysed with flow cytometry (figure 3g) demonstrating prescriptive control over the relative mean concentration of ssDNAPEGDPPE on cell surfaces by the addition of meoPEGDPPE during modification. As with flow cytometry-based fluorescence intensity, the mean molecular surface concentration correlates inversely with the meoPEGDPPE : FAM–SeqAPEGDPPE molar ratio (figure 3h).

Sampling from the same populations of cells used for flow cytometry and fluorescence spectroscopy analyses, we also systematically gathered confocal images of cell cross-sections. Images were processed semi-automatically by first manually identifying the centres of each cell prior to a polar-Cartesian transform and an automated pixel alignment of the resulting linear membrane image. The vertically aligned membrane images were each collapsed into 1D intensity profiles that could be integrated for each cell analysed (electronic supplementary material, figure S5). The polar-Cartesian transformations were non-conservative in that the density of pixel brightness rather than total pixel values was preserved, thus integrated membrane profile values correspond to intensity densities which, in turn, correlated positively with both mean peak flow cytometry intensities (figure 3i) and the molecular surface concentrations computed from fluorescence spectroscopy (figure 3j).

3.4. Fluorescence-based inference of hybridization at doublet interfaces

In addition to demonstrating prescriptive meoPEGDPPE-based control over the molecular concentration of ssDNAPEGDPPE on cells prior to doublet formation, the agreement among the three fluorescence-based characterization methods allows us to make inferences about the molecular concentration of hybridizing molecules based on the images of membrane profiles. To investigate the relationship between adhesion, as we have computed it in our mechanical–geometric doublet model and the molecular surface concentration of ssDNAPEGDPPE at the cell–cell interface, we processed doublet images (n = 50 per modification condition) using the manually measured cell centre point data and radii to obtain automatically isolated images of the cell–cell interfacial profile (figure 4a). These images were then converted to vertically aligned images. Observing the distribution of intensity along the length of these images, intensities were attenuated within 1 μm from the perimeter of the interface, but otherwise apparently uniform throughout the centre region i.e. standard deviations less than 10% in pixel intensity for most cross-sections. The aligned images were collapsed into 1D profiles (figure 4b) and integrated to obtain a single mass intensity value for each doublet interface. Across all tested meoPEGDPPE : FAM–SeqAPEGDPPE ratios, we observed a positive correlation between individual doublet mass intensities and their corresponding ${\overline{A}}_{\mathrm{contact}}$ (figure 4c; electronic supplementary material, figure S6). Importantly, when we analysed cytochalasin D-treated doublets, we observed a similar positive correlation between mass intensities and their corresponding ${\overline{A}}_{\mathrm{contact}}$ across the range of doublets analysed, however with an apparently lower modulus than the non-cytochalasin D-treated doublets.

We converted integrated intensity profiles to fluorescence intensity-based surface densities by dividing them by the interface diameter a_ij while assuming a constant confocal microscope sectioning depth. When the geometrically determined mean adhesion values for different meoPEGDPPE : ssDNAPEGDPPE ratios are plotted against intensity densities, a nonlinear trend emerges (figure 4d), and the plot of ${\overline{A}}_{\mathrm{contact}}$ versus intensity density similarly reveals a nonlinear relationship. These relationships plotted as log-log plots suggested power-law scaling (electronic supplementary material, figure S7) that could not be accounted for by camera saturation (electronic supplementary material, figure S8). The intensity density of the cell–cell interface should correspond to the molecular concentration of fluorescein and thus to the concentration of oligonucleotides at the interface; however, because doublet interfaces exhibited intensity densities spanning a range past the upper limit of those taken from the images of single cell membrane profiles (figure 4e), we were unable to construct a reliable interpolative standard to infer the molecular concentration of FAM-SeqAPEGDPPE at doublet interfaces and instead applied a less rigorous upward extrapolation (electronic supplementary material, figure S9).

4. Discussion and conclusion

We used lipid-conjugated oligonucleotides to tune average intercellular adhesion in the absence of cadherins. A tomographic reconstruction of the 3D structure of a ssDNAPEGDPPE-co-adhered cell doublet revealed that fluorescently labelled ssDNA was distributed throughout the interface between the two cells (figure 1b) in contrast, for example, to the annular distribution of cadherins at doublet interfaces observed by Engl et al. [15]. This difference marks an aspect of ssDNAPEGDPPE-mediated adhesion which departs from adhesion modes that interface with the internal cytoskeleton and participate directly in an active force generating/sensing feedback loop. We examine ssDNAPEGDPPE doublets at a pseudo-steady state, though previous studies suggest changes would take place over longer timescales of hours such as the internalization of oligos [27]. The timescale of DNA hybridization is rapid, with the contact-formation phase occurring within seconds. Our system is also likely to involve actomyosin feedback in the form of cortex remodelling. This has been shown to occur on timescales of minutes [43,44] in response to cell shape change or external deformation. Our system, therefore, while cadherin-free, is nevertheless not completely orthogonal to cytoskeletal dynamics.

Our model describes the space of geometric configurations (represented by ${\overline{A}}_{\mathrm{contact}}$) in terms of the Young’s modulus and adhesion, postulating effective independence (figure 2d). We hypothesized that modulation of one and/or the other could permit navigation over the configuration space. To explore this principle experimentally, we used a non-hybridizing meoPEGDPPE during ssDNAPEGDPPE modification to reduce the adhesion of cells and observed that the interfacial contact area decreased compared to doublets without meoPEGDPPE. We observed conversely that cells treated with cortical actin-inhibiting cytochalasin D exhibited large interfacial contact areas despite undergoing analogous ssDNAPEGDPPE modification as untreated cells (figure 2a–c). By preparing complementarily modified cells with different meoPEGDPPE : ssDNAPEGDPPE ratios, we were able to systematically shift the distributions of interfacial contact areas determined from microscopy image analyses (figure 2e) and converting the measurements of doublet geometry into mean adhesion estimates via equation (2.6) shows that systematic shifts in doublet geometry correspond to analogous shifts in the adhesion (figure 2f). For simplicity, we assume a constant Young’s modulus independent of cortical rearrangement, conceding that it is likely to be affected in the case of highly deformed doublets. Nevertheless, we see that the meoPEGDPPE relative concentration enables prescriptive adjustment of cell–cell adhesion (figure 2g).

Our results apply to an ongoing discussion in cytoskeletal biology over what extent spontaneous separation of different cell types within tissues and the interfacial tension of multicellular domains are a result of differential adhesive compatibility analogous to phase separation of polar and non-polar liquids [14,45–48] relative to the cortical stiffness of cells [49,50]. A confounding factor comes from the annular actin reorganization in cadherin–cadherin-cohering cell doublets, which concentrates at the circumference of the cell–cell interface where cadherins also accumulate [15,51]. Without direct coupling to the internal cytoskeleton in our case, actin remodelling is likely restricted to that which occurs in response to compressive deformation. The treatment of doublets with cytochalasin D under normal ssDNAPEGDPPE modification conditions and subsequent increased mean interfacial contact area demonstrates that the doublet geometry is influenced by the elastic mechanics of the cells. In a direct analogy to tissue surface tension, characterized by a minimization of the total exposed surface area multicellular aggregates to total aggregate volume ratio, the surface area to volume ratio of the cell doublet decreases with increasing interfacial contact area, indicating a balance of cortical stiffness and adhesion in agreement with current models [52–54].

In sum, we report on the establishment of a model to study cell–cell adhesion with a minimization of the active contribution of cohesion-inducing proteins like cadherins. By tuning the ratio of PEG-only and oligonucleotide-bearing lipid conjugates or by the introduction of stiffness-influencing pharmacological factors, it was possible to affect the size of the contact area of co-adhering cell doublets. We use a mechanical model to describe the differences in contact area under different parameters. The system could be of value to studies that seek to examine the impact of adhesion without a major confounding influence of other cytoskeletal factors.

Data accessibility

This article has no data.

Authors' contributions

I.T.H., Y.A. and H.I. conceived the research. I.T.H. carried out the research. I.T.H., Y.A. and H.I. wrote the manuscript.

Competing interests

We declare we have no competing interests.

Funding

This study was supported by a Grant-in-Aid for Scientific Research on Innovative Areas ‘Nanomedicine Molecular Science’ (grant no. 2306) and the Monbukagakusho Scholarship for graduate studies from the Ministry of Education, Culture, Sports, Science, and Technology (MEXT) of Japan.