Imaging Membrane Order and Dynamic Interactions in Living Cells with a DNA Zipper Probe

Abstract

The cell membrane is a dynamic and heterogeneous structure composed of distinct sub-compartments. Within these compartments, preferential interactions occur among various lipids and proteins. Currently, it is still challenging to image these short-lived membrane complexes, especially in living cells. In this work, we present a DNA-based probe, termed “DNA Zipper”, which allows the membrane order and pattern of transient interactions to be imaged in living cells using standard fluorescence microscopes. By fine-tuning the length and binding affinity of DNA duplex, these probes can precisely extend the duration of membrane lipid interactions via dynamic DNA hybridization. The correlation between membrane order and the activation of T-cell receptor signalling has also been studied. These programmable DNA probes function after a brief cell incubation, which can be easily adapted to study lipid interactions and membrane order during different membrane signalling events.

Graphical Abstract

Cholesterol-modified programmable DNA probes are developed to image transient interactions and membrane order in live cell membranes. The membrane order during the activation of T-cell receptor signalling has also been measured.

Introduction

The plasma membrane is an intriguingly dynamic and complex structure that regulates cellular interactions with the environment. A large number of cell signaling processes occur or start at the plasma membranes, and many of these signaling events are regulated by specific interactions among membrane lipids and proteins.^[1] Preferential interactions of these membrane components can promote lateral heterogeneity and segregation in the membrane plane.^[2] Such heterogeneity further influences the distribution and stability of membrane receptor clusters and other complexes to facilitate localized signaling.^[2,3] For example, the activation of T-cell receptor (TCR) signaling is regulated by the formation of membrane lipid order.^[4–7] TCRs can form oligomers within these membrane phases, which is critical in improving their selectivity towards antigenic ligands. Disruption of the lipid composition of T-cell membranes, especially cholesterols and sphingolipids, can result in inflammation and autoimmune diseases.^[4,5,8,9]

Both stable (seconds–hours) and transient (μs–ms range) molecular interactions occur in the cell membranes.^[10] Compared to strong/stable interactions, the formation of weak/transient membrane complexes are often more difficult to measure.^[11] Furthermore, traditional methods for studying membrane complexes, such as NMR and mass spectrometry, are normally used to analyze in vitro prepared samples or cell membrane extracts.^[10,12–15] Transient membrane complexes can be easily disrupted during these isolation and purification processes. As a result, the existence of transient membrane interactions can often be overlooked or misevaluated. Thus, our understanding of the mechanisms and outcomes of these short-lived membrane interactions are still limited.

To measure transient interactions in living cell membranes, several fluorescence-based approaches have been developed, including fluorescence correlation spectroscopy (FCS) and cross-correlation spectroscopy (FCCS),^[16,17] super-resolution microscopy,^[18–21] single-molecule tracking,^[22,23] and Förster resonance energy transfer (FRET).^[24,25] These methods can provide valuable information about the temporal/spatial distributions of membrane interactions on live cell surfaces. Among these methods, FRET microscopy is particularly powerful because it enables real-time monitoring of molecular interactions across the entire cell membrane, and also doesn’t require costly instrumentation or complex data analysis.^[26]

To apply the FRET technique to the study of membrane interactions, a pair of donor/acceptor fluorophores or fluorescent proteins is often used to label the target membrane partners, respectively. Once the labelled targets are within a distance of ~1–10 nm, a decrease in the donor emission and an increase in the acceptor fluorescence occurs (Figure 1a). As a result, the FRET signal can be used to monitor membrane interactions in real-time. While FRET microscopy has been widely used for studying relatively stable molecular interactions, due to low photon emission and limited signal-to-noise ratio, it is still difficult to image the formation of short-lived membrane complexes, especially between lipid molecules.^[10] In this study, we aim to develop a type of DNA probes that enables FRET-based imaging of membrane transient interactions on live cells.

Figure 1.

Design and in vitro characterization of the DNA Zipper. a) Schematic representation of traditional FRET sensor (top) and DNA Zipper (bottom) for membrane interaction studies. The duration (1/k_off) and FRET efficiency of membrane interactions can be increased by incorporating “n” base pairs in the DNA Zipper design. b) The FRET efficiency of DNA Zipper was imaged on a DLPC-based supported lipid bilayer. 50 nM of each P_n strand was first incubated with the lipid bilayer for 15 min, and then 0–500 nM of the P* strand was added for 15 min before imaging. Samples were excited at 488 nm and the fluorescence emission of Atto 488 (I₅₂₀) and Atto 647N (I₆₆₀) was collected at 520 and 660 nm, respectively. I₅₂₀ denotes the donor emission detected using a 511/20 bandpass filter, while I₆₆₀ denotes the acceptor emission detected using a 685/40 bandpass filter. Pseudo-color I₆₆₀/I₅₂₀ ratiometric images were also shown. Scale bar, 20 μm. c) The correlation between FRET efficiency and membrane concentration of P* was determined for each pair of P_n/P*. The FRET efficiency was calculated based on P*-induced fluorescence intensity changes in the donor (I₅₂₀) channel. Shown are the mean and standard error of the mean (SEM) values from 20 lipid bilayer areas in each case.

DNA-based probes, with exceptional properties of programmable self-assembly and precise hybridization, have been widely used for in vitro and intracellular detections.^[27,28] Since the recent discovery that synthetic lipid-DNA conjugates can spontaneously incorporate into cell membranes, these DNA-based probes have been further applied for cell membrane engineering and analysis.^[29–33] As an example, we recently developed a DNA strand displacement-based strategy to convert individual transient membrane lipid–lipid interactions into cumulative fluorescence signal changes.^[34] Even though this approach cannot be used to directly monitor the spatial distributions or the dynamic variations of membrane molecular interactions, the existence of some previously untraceable membrane order interactions were able to be validated in living cells.

In this study, we report the development of a “next-generation” DNA probe, called “DNA Zipper” to directly monitor lipid-mediated membrane interactions in living cells. Compared to traditional FRET microscopy, the membrane target partners of interest are not directly conjugated to donor/acceptor fluorophores in the DNA Zipper, but instead, through short complementary DNA strands (Figure 1a). The DNA Zipper functions by extending the duration of membrane transient interactions, in a predefined and predictable pattern as a result of DNA hybridization, and thus converting transient and weak membrane interaction signals to easily detectable fluorescence readouts. After testing a series of DNA probes with different hybridization stability and kinetics, our results indicated that the optimized DNA Zippers can be used for studying the heterogeneous patterns and dynamic changes of membrane interactions on live cell surfaces, e.g., during the activation of T-cell receptor signaling.

Results and Discussion

Design and In Vitro Characterization of DNA Zippers

To test the DNA Zipper concept, we chose to study membrane cholesterol-mediated probe interactions. Cholesterol was chosen because it plays a key role in the formation of membrane lipid order and can also regulate the lateral organization of various lipids and proteins.^[2,35] These dynamic cholesterol-based interactions are still difficult to visualize in live cell membranes.^[36]

To develop DNA Zipper probes for membrane analysis, we first synthesized two types of cholesterol-modified 15-nucleotide (nt)-long DNA oligonucleotides, denoted as the donor (P_n) and acceptor (P*) strand, respectively. A FRET donor dye (Atto 488) was linked to the 5’-end of the P_n strand, and a corresponding acceptor fluorophore (Atto 647N) was conjugated to the 3’-end of the P* strand, with cholesterol attached at the other end of each strand (Table S1 and Figure S1). The P_n and P* strands were designed to partially hybridize with each other through “n” base pairs (n= 6–10) (Figure 1a). As a control, we also prepared two other donor strands that will either not bind (P₀) or fully hybridize (P₁₅) with the acceptor P* strand. All of these strands have been designed to exhibit no other secondary structures (Figure S2).

We first wanted to study if the FRET efficiency of the P_n/P* complex can be regulated based on the length of DNA duplex. Indeed, with an increasing number of base pairs between P_n and P*, a gradual decrease in the donor fluorescence peak (λ_ex= 488 nm, λ_em= 520 nm) and an increase in the acceptor fluorescence (λ_ex= 488 nm, λ_em= 660 nm) was observed (Figure S3). Longer DNA hybridization results in a stronger binding and higher FRET efficiency in solution.

To test if DNA hybridization-regulated tuning of the FRET efficiency can also be observed on a two-dimensional membrane surface, we prepared a 1,2-dilauroyl-sn-glycero-3-phosphocholine (DLPC) symmetric lipid bilayer. Both the donor and acceptor DNA probes can be efficiently anchored onto this supported lipid bilayer (Figure S4). As shown in Figure 1b and Figure S5, the FRET signals on the supported lipid bilayer can be precisely regulated by controlling the surface densities of the P* strand. The more P* in the bilayer, the higher the FRET signal. Also as expected, the membrane FRET efficiency can be tuned by adjusting the length of DNA duplex within the P_n/P* conjugates (Figure 1b).

The correlation between the membrane FRET efficiency and the concentration of P* for each pair of the P_n/P* conjugates was further determined (Figure 1c). Here, the FRET efficiency (E) is calculated as E= 1 – (I_Pn,P*/I_Pn), where I_Pn,P* and I_Pn stand for the membrane averaged donor P_n fluorescence intensity in the presence and absence of the acceptor P* strand, respectively. When the number of base pairs increased from P₆/P* to P₁₅/P*, the apparent membrane dissociation constant K_D gradually decreased from 2.8×10⁻⁷ μmol/mm² to 6.7×10⁻⁸ μmol/mm² (Table S2). While in contrast, no binding event or a K_D value can be determined for the P₀/P* pair. All these results indicated that by adjusting the DNA duplex length, we can fine-tune the FRET efficiency of the DNA Zipper both in solution and on a membrane surface.

Membrane Diffusion and Phase-Partitioning of DNA Zippers

Before using the DNA Zipper to study dynamic membrane interactions, we wondered if the membrane lateral diffusion of the cholesterol-DNA conjugates would be significantly different from that of fluorophore-labeled cholesterols. We first determined the diffusion coefficient of P₁₀ in the DLPC-based supported lipid bilayer membranes using a fluorescence recovery after photobleaching (FRAP) approach.^[37] 2,3-dipyrrometheneboron difluoride-labeled cholesterol (TF-Chol), a commonly used fluorescent cholesterol compound, was used as a control.^[38,39] In the DLPC supported lipid bilayers, TF-Chol and P₁₀ exhibit very similar diffusion coefficient, 0.046±0.003 μm²/s vs. 0.045±0.004 μm²/s (Figure S6).

We further used FRAP to measure the diffusion coefficients of P₁₀ and TF-Chol in the membrane of a model Madin-Darby canine kidney (MDCK) cell line. DNA probes can efficiently insert onto these MDCK cell membranes, with over 90% of the probe inserted during an ~8 min incubation (Figure S7). Again, a quite similar diffusion coefficient was observed for the P₁₀ (0.20±0.02 μm²/s) and TF-Chol (0.21±0.02 μm²/s) probe (Figure S8). Indeed, DNA-modified cholesterols exhibit some similar membrane diffusivity as fluorophore-labeled cholesterols.

We next asked whether the DNA probe and TF-Chol exhibit similar membrane partitioning. Membrane order can be generally classified as liquid-ordered (L_o) and liquid-disordered (L_d) phases. With the lack of actin cytoskeleton and easy phase separation, giant plasma membrane vesicles (GPMVs) are often used as a testing platform of membrane order.^[40] After extracting GPMVs from the MDCK cells, to distinguish the L_o and L_d phases in the GPMVs, we added cholera toxin B subunit (CT-B) to specifically stain the L_o phases.^[41] TF-Chol and the DNA probe exhibit similar membrane phase partition coefficients (P_o/d), i.e., 2.25±0.38 and 2.04±0.44, respectively (Figure S9). Here, the partition coefficient is defined as the ratio of probe concentration in the L_o vs. L_d phases. All these data indicated that modification of DNA probes will not significantly alter the membrane lateral diffusion or phase partitioning of fluorophore-labeled cholesterols.

Imaging DNA Zipper Interactions in Live Cell Membranes

Our next goal was to test if dynamic DNA Zipper interactions could be imaged in live MDCK cell membranes. Considering cell-to-cell variations in the membrane probe densities, neither donor fluorescence (I₅₂₀, i.e., λ_ex= 488 nm, λ_em= 520 nm) nor the FRET signal (I₆₆₀, i.e., λ_ex= 488 nm, λ_em= 660 nm) itself is reliable for measuring membrane distributions of DNA Zipper interactions. We realized that there is a strong linear correlation between cellular I₅₂₀ and I₆₆₀ signals (Figure S10). As a result, the I₆₆₀/I₅₂₀ ratiometric signal can be independent of membrane probe density differences and thus will be used for later analyses.

We first tested if the DNA Zipper can dynamically associate and dissociate in live-cell membranes. For this purpose, after incubating 50 nM of the P₁₀ probe with the MDCK cells and washing away excess probes, 100 nM of P* was added. Very efficient P₁₀/P* hybridization and a dramatic increase in the membrane I₆₆₀/I₅₂₀ ratiometric signal was observed within 10 min (Figure S11). After another 30 min incubation, a P*_CS strand was added, which is fully complementary to P* so that it can displace and free P₁₀ from the original P₁₀/P* duplex. As expected, a rapid dissociation of P₁₀/P* (decreased I₆₆₀/I₅₂₀ signal) was shown in only several minutes (Figure S11). To further validate our data, fluorescence lifetime imaging (FLIM) was also performed in this test, and very consistent results were shown (Figure S11). These results indicated that the DNA Zipper can indeed rapidly hybridize and dehybridize in the cell membranes, in a highly controllable configuration.

Next, regulation of the FRET efficiency of DNA Zippers was studied by changing duplex lengths and acceptor concentrations in MDCK cell membranes. Similar to the lipid bilayer results (Figure 1b), the FRET efficiency of the DNA Zippers can be fine-tuned based on the number of base pairs, as well as the membrane densities of P* (Figure 2a and Figure S12). The apparent membrane dissociation constant K_D values gradually decreased from 4.9×10⁻⁸ μmol/mm² (P₆/P*) to 3.1×10⁻⁹ μmol/mm² (P₁₅/P*) as the length of DNA duplex increased (Figure 2b). Some similar result was also observed in FLIM-based analysis (Figure S13). Interestingly, these obtained K_D values are about an order of magnitude lower than that in the DLPC lipid bilayers (6.7×10⁻⁸–2.8×10⁻⁷ μmol/mm²). This may be due to the much faster probe diffusion along the cell membranes (Figures S5 and S7), which will increase the probability of probe encounters (k_on).

Figure 2.

Characterizing DNA Zipper interactions in live cell membrane. a) 50 nM of each P_n strand was first incubated with the MDCK cells for 20 min, and then 0–500 nM of the P* strand was added for 15 min before imaging. Samples were excited at 488 nm and the fluorescence emission of Atto 488 (I₅₂₀) and Atto 647N (I₆₆₀) was collected with a 511/20 and 685/40 filter, respectively. Pseudo-color I₆₆₀/I₅₂₀ ratiometric images were also shown. Scale bar, 20 μm. b) The correlation between FRET efficiency and cell membrane concentration of P* was determined for each pair of P_n/P*. The FRET efficiency was calculated based on P*-induced fluorescence intensity changes in the donor (I₅₂₀) channel. Shown are the mean and SEM values from 20 cell membranes in each case. c) After extracting GPMVs from MDCK cells and stained with 2.5 μg/ml cholera toxin B subunit (CT-B) for 30 min at 4°C, 5 nM of P₇/P* was added and incubated for 30 min before imaging. By comparing the membrane distribution of the CT-B fluorescence and the I₆₆₀/I₅₂₀ ratio, the percentage of L_o phase-associated P₇/P* signals were determined. Scale bar, 5 μm. Shown are the mean and SEM values obtained on the membranes of 20 GPMVs.

Our next goal is to determine if the dynamic DNA Zipper interactions can be imaged in live cell membranes. The effect of methyl-β-cyclodextrin (MβCD) treatment on the DNA Zipper signals was chosen. MβCD is a widely used lipid order-disrupting agent that can selectively extract cholesterols from cell membranes.^[42] The depletion of cholesterols results in a decreased level of membrane L_o phases. Indeed, after adding MβCD, the percentage of membrane L_o phases on the GPMVs was significantly reduced (Figure S14). In these tightly packed L_o phases, the high local concentration of cholesterols facilitates more effective encounters among DNA Zippers (Figure 2c), and as a result, the addition of MβCD can be used to reduce membrane interactions and FRET signals of DNA Zippers. After treating MDCK cells with 0.1–20 mM MβCD, a concentration-dependent decrease in the I₆₆₀/I₅₂₀ ratiometric signal of a P₇/P* DNA Zipper probe can be observed (Figure S14).

We next asked what duplex lengths of the DNA Zipper can be applied to image MβCD-induced changes in the membrane DNA Zipper interactions. For this purpose, MDCK cells were first treated with 10 mM MβCD for 20 min, and then each pair of DNA Zipper was added. By comparing their cell membrane ratiometric signal before and after the MβCD treatment, a significantly reduced FRET efficiency was observed for the P₆/P*, P₇/P*, and P₈/P* probes (Figure 3a and Figure S15). While P₉/P* and P₁₀/P* exhibited higher FRET efficiency, there was no change in their I₆₆₀/I₅₂₀ ratios, which is likely because these DNA hybridizations are too stable to be disrupted by the MβCD treatment (Figure 3b). Based on these results, DNA Zippers with 6–8 complementary bases should be suitable for measuring dynamic changes in cell membrane interactions.

Figure 3.

Imaging MβCD-induced disruption of membrane DNA Zipper interactions. a) MDCK cells were first treated with 10 mM MβCD for 30 min, and then 50 nM of each pair of P_n/P* was added for 15 min before imaging. Cells without the MβCD pre-treatment were used as the control in each case. Pseudo-color I₆₆₀/I₅₂₀ ratiometric images were shown on samples irradiated at 488 nm. Scale bar, 20 μm. b) The average I₆₆₀/I₅₂₀ ratiometric signals on individual MDCK cell membranes in the presence or absence of the MβCD pre-treatment as measured at 25°C. c) The effect of temperature on the performance of DNA Zippers on MDCK cell membranes. Shown are the mean and SEM values from at least 30 cell membranes in each case. **p< 0.01, ***p< 0.001 in two-tailed Student’s t-test; ns, not significant.

Note that all the above experiments were performed at room temperature (25°C). To study the effect of temperature on the performance of DNA Zippers, we also measured MβCD-induced membrane signal changes of each P_n/P* probe at 4°C and 37°C (Figure 3c and Figure S16). Again, our results indicated that P₆/P*, P₇/P*, and P₈/P* probes can be applied to study dynamic membrane interactions at 37°C. In contrast, because the P₈/P* hybridization is too stable at 4°C, we can only use P₆/P* and P₇/P* to image MβCD-induced reduction in membrane DNA Zipper interactions at low temperature.

Distinguishing Membrane Order with DNA Zippers

Due to the relatively high probability of DNA Zipper interactions in L_o phases (Figure S9a), we wondered if the DNA Zipper could be used to distinguish and selectively image membrane liquid ordered phases. To test this hypothesis, in GPMV membranes the fluorescence colocalization of CT-B (a L_o phase-staining dye) with the I₆₆₀/I₅₂₀ ratiometric signal of the P₇/P* DNA Zipper was studied. Quite encouragingly, there is a very strong correlation between the membrane staining patterns of these signals (Figure 2c). It is worth mentioning that compared to single-stranded DNA-modified cholesterols, DNA double strand-modified cholesterol probes exhibit even higher tendency towards L_o phase partitioning (Figure S9). P₇/P* can be thus used to identify membrane order in the GPMVs.

We next wanted to know if the DNA Zipper could also be used to study membrane order in living cells. Similar to the above-mentioned MβCD treatment, we decided to further test the effect of cholesterol sulfate, another agent that is known to prevent membrane association of cholesterols and thus potentially disrupt lipid order.^[43] The P₇/P* probe was chosen as it showed an optimal response to MβCD-induced lipid order disruption (Figure 4). Indeed, after adding 100 μM cholesterol sulfate to the MDCK cells, a significantly reduced I₆₆₀/I₅₂₀ membrane ratiometric signal was observed, consistent with that of MβCD disruption (Figures 4a, 4b and Figure S17).

Figure 4.

Studying membrane order with the DNA Zipper. a) MDCK cells were first treated with 10 mM MβCD, 100 μM cholesterol sulfate, a mixture of 10 mM MβCD and 1 mM cholesterol (cholesterol supplementation), or a mixture of 40 mM MαCD and 1.5 mM sphingomyelin (sphingomyelin supplementation), respectively for 30 min, and then 50 nM of P₇/P* was added for 15 min before imaging. Cells without these treatments were used as a control. Representative pseudo-color I₆₆₀/I₅₂₀ ratiometric images were shown on samples irradiated by 488 laser. Scale bar, 20 μm. Shown were also the corresponding percentages of weak, moderate, and strong DNA Zipper interactions in the MDCK cell membranes, which was determined based on the I₆₆₀/I₅₂₀ ratiometric signal. b) The average I₆₆₀/I₅₂₀ ratiometric signals on individual MDCK cell in the presence or absence of treatment with each lipid order inducing/disrupting agent. ***p< 0.001 in two-tailed Student’s t-test. Shown are the mean and SEM values from at least 20 cells in each case.

We also studied the effect of cholesterol and sphingomyelin supplementation on the membrane DNA Zipper interactions. Both of these treatments will promote the formation of membrane L_o phases (Figure S14).^[44,45] As expected, increased cell membrane I₆₆₀/I₅₂₀ signals from the P₇/P* probe were observed, indicating the existence of more frequent membrane DNA Zipper interactions (Figures 4a, 4b and S17). Interestingly, our results indicated that the supplementation of cholesterols reduced the cell membrane diffusion coefficient of the P₇ probe (Figure S14). We thus believe both the membrane local concentration and diffusion rate of the cholesterol-DNA conjugates will regulate the interaction efficiency of the DNA Zipper.

Our next goal is to map the spatial distributions of DNA Zipper interactions on the MDCK cell membranes. To this end, based on the mean intensity and associated variances among individual cells (See SI Methods), we categorized the I₆₆₀/I₅₂₀ ratiometric signal in the plane of cell membranes into three groups: weak, moderate, and strong. Our results indicated that after treating the MDCK cells with 10 mM MβCD or 100 μM cholesterol sulfate, the percentage of cell membrane area with low DNA Zipper interaction signals increased from 63% to 85% and 97%, respectively (Figure 4a). While only 5% and 1% of the cell membranes exhibited signals indicative of frequent DNA Zipper interactions, as compared to 17% before the disruption. On the other hand, under the same experimental condition, the supplementation of excess cholesterol and sphingomyelin significantly increased the percentage of regions with frequent DNA Zipper interactions to 40% and 43%, respectively (Figure 4a). It is worth mentioning that since arbitrary threshold values were used, the percentage of frequent/strong DNA Zipper interactions doesn’t necessarily report on the proportion of membrane L_o phases, but these values can be good indicators of changes in the membrane order distributions in living cells.

It is quite interesting that the formation/deformation and lateral migration of segmented cholesterol interactions can also be monitored in the GPMVs using the DNA Zipper. For example, as shown in Figure S18, the deformation of segmented membrane regions with frequent DNA Zipper interactions happens within ~3 s, indicating a dynamic nature of these membrane interactions and lipid order.

Correlation of Membrane Order with the Activation of T-Cell Receptor (TCR) Signaling

After demonstrating the utility of the DNA Zipper in imaging molecular interactions and membrane order in living cells, we next asked if the DNA Zipper could be further applied to study the roles of membrane order at the onset of T-cell signaling. The existence of lipid order plays an important role in the formation of TCR signaling complexes.^{[4–7,46–50]} Membrane order can act as a platform to allow TCRs to interact with kinases and trigger signaling cascades that promote T-cell proliferation and differentiation.^[50]

To test if the DNA Zipper can be used to investigate lipid order formation due to the onset of TCR signaling, we first studied the hybridization efficiency of DNA probes in a model T-cell membrane, the CD4⁺ Jurkat E6–1 cell line, which has been widely used for studying TCR signaling.^[51,52] Using P₇ as an example, efficient and rapid Jurkat cell membrane insertion of the DNA probe can be observed in ~20 min (Figure S19). Similar to what was observed in the MDCK cell membranes (3.1×10⁻⁹–4.9×10⁻⁸ μmol/mm²), by gradually increasing the length of DNA duplex from P₆/P* to P₁₅/P*, the FRET efficiency of the DNA Zipper was also correspondingly enhanced in these Jurkat cell membranes (1.1×10⁻⁸–3.9×10⁻⁸ μmol/mm²) (Table S2, Figures S20 and S21).

We also wondered if the membrane insertions of DNA Zippers would influence the stimulation of TCR signaling. Here, a cell permeant Ca²⁺ indicator, X-Rhod-1, was used to track T-cell stimulation-induced increases in intracellular calcium levels.^[52] After adding a mixture of anti-CD3 antibody (triggering TCR clustering)^[53] and anti-CD28 antibody (co-stimulator),^[54] a very similar level of the X-Rhod-1 fluorescence enhancement (~1.5-fold) was observed in the presence or absence of different DNA Zipper probes (Figure S22). As a control, the addition of anti-CD28 antibody itself, without engaging with the TCR complex, didn’t induce any noticeable Ca²⁺ signals in all these experiments. These results indicated that the membrane insertion of DNA probes will not obviously affect the TCR signaling in this Jurkat cell line.

We next asked how the membrane signals from DNA Zippers of different duplex lengths would be changed during the stimulation of TCR signaling. After pre-incubating the Jurkat cells with 1 μM X-Rhod-1 and 50 nM P_n and 100 nM P* probes, excess probes were removed and a mixture of anti-CD3 and anti-CD28 antibodies was added to co-stimulate the T-cells. As a control, Jurkat cells were treated with either buffer or only the anti-CD28 antibody. The results indicated that similar to what we observed in the MDCK cell membranes, P₆/P*, P₇/P*, and P₈/P* are the optimal DNA Zippers and all these probes exhibit a significant increase in the I₆₆₀/I₅₂₀ signal upon an antibody-induced T-cell stimulation (Figure 5a and Figure S23).

Figure 5.

DNA Zipper signal changes during the stimulation of TCR signalling. a) 50 nM P_n and 100 nM P* probes were incubated with CD4⁺ Jurkat cells for 20 min. Shown are the average I₆₆₀/I₅₂₀ ratiometric signals on individual cell membranes before and 2 min after adding a mixture of 10 μg/mL of anti-CD3/anti-CD28 antibodies. b) Jurkat cells were first treated with 0–250 μM of cholesterol sulfate for 2 h, and then 50 nM of P₇ and 100 nM P* probes were added and incubated for 20 min. The average I₆₆₀/I₅₂₀ ratiometric signals on individual cell membranes were shown before and 2 min after the treatment with 10 μg/mL of anti-CD3/anti-CD28 antibodies. c) These Jurkat cells were first treated with 0–5.0 mM of MβCD for 30 min, and then 50 nM of P₇ and 100 nM P* probes were added and incubated for 20 min. The average I₆₆₀/I₅₂₀ ratiometric signals on individual cell membranes were shown before and 2 min after the treatment with 10 μg/mL of anti-CD3/anti-CD28 antibodies. d) The percentage of weak, moderate, and strong DNA Zipper interactions in the Jurkat cell membranes as determined based on the I₆₆₀/I₅₂₀ ratiometric signal of P₇/P* before (resting) and 2 min after (activated) adding a mixture of 10 μg/mL of anti-CD3/anti-CD28 antibodies. Shown are the mean and SEM values from at least 20 cells in each case. *p< 0.05, **p< 0.01, ***p< 0.001 in two-tailed Student’s t-test; ns, not significant.

It has been suggested that disruption of membrane order will potentially have a negative impact on T-cell signaling.^[55,56] Our next goal was to test if the observed changes in the membrane FRET signals are indeed due to the formation or enhanced stabilization of membrane order during TCR stimulation. For this purpose, we studied the effect of MβCD- and cholesterol sulfate-induced membrane order modulation on TCR stimulation and the corresponding DNA Zipper signals. Indeed, after treating the Jurkat cells with increasing concentrations of MβCD or cholesterol sulfate, the efficiency of anti-CD3/anti-CD28-incuded calcium flux was dramatically reduced (Figure S24). Also as expected from the effect of lipid order disruption on reducing membrane DNA Zipper interactions, the cell membrane I₆₆₀/I₅₂₀ ratiometric signal of the DNA Zipper (P₇/P* was chosen here) was gradually decreased after treating with increasing amounts of MβCD or cholesterol sulfate (Figures 5b and 5c). Meanwhile, after adding high concentrations of MβCD or cholesterol sulfate, no difference in the DNA Zipper signal was shown before or after the co-stimulation with a mixture of anti-CD3/anti-CD28 antibodies. This result is consistent with the fact that after significantly disrupting membrane order, the addition of anti-CD3/anti-CD28 antibodies is deficient in triggering TCR signaling.^[57,58] All these data suggested that the membrane DNA Zipper signal can indeed be used to study membrane order and its effects during TCR signaling.

The membrane residence of the TCR/CD3 complex in cholesterol/sphingolipid-rich L_o phases is a key hypothesis to explain the role of membrane order in regulating TCR activation.^[59] We next aimed to study if the membrane distributions of DNA Zipper interactions are indeed directly correlated with that of the TCR/CD3 complex. The changes in membrane distributions of strong, moderate, and weak DNA Zipper interactions were first mapped during TCR stimulation. After adding anti-CD3/anti-CD28 antibodies, a significant increase (from 16% to 29%) was shown in the percentage of membrane regions with frequent DNA Zipper encounters (Figure 5d), which indicated an enlargement of membrane L_o phase areas. To study the membrane colocalization of DNA Zipper interactions with the TCR/CD3 complex, the DNA Zipper probe was imaged together with a Pacific Blue^™ dye-conjugated anti-CD3 antibody, which can be used for simultaneous co-stimulating TCR and imaging membrane TCR/CD3 clusters. Indeed, a strong positive correlation was observed between the cell membrane CD3 signals and the I₆₆₀/I₅₂₀ ratiometric signals of P₇/P* (Figure 6a).

Figure 6.

Temporal and spatial correlations of the DNA Zipper membrane interaction patterns with the TCR signaling. a) Correlation between the I₆₆₀/I₅₂₀ ratiometric signal of P₇/P* and the membrane Pacific Blue^™ channel intensity (anti-CD3 antibody) on individual CD4⁺ Jurkat cell membranes. The correlation was studied at 2 min after the co-stimulation with fluorescent anti-CD3 and anti-CD28 antibodies. b) The correlation between the extent of DNA Zipper interactions and the membrane Pacific Blue^™ channel intensity (anti-CD3 antibody) at different time point after adding a mixture of 10 μg/mL of anti-CD3/anti-CD28 antibodies. The percentage of weak, moderate, and strong DNA zipper interactions in individual Jurkat cell membranes was determined based on the I₆₆₀/I₅₂₀ ratiometric signal of P₇/P*. c) At zero second, 10 μg/mL of anti-CD3/anti-CD28 antibodies were added to stimulate TCR signaling. Cell membrane I₆₆₀/I₅₂₀ ratiometric signals (as measured with P₇/P*) were monitored over 5 min after the stimulation. As a control, cell membrane DNA Zipper signals were also monitored after adding only 10 μg/mL of anti-CD28 antibody or without any antibody stimulation (control). d) For the same cells shown in the panel (c) the corresponding cellular accumulation of Ca²⁺ ions was monitored based on the fluorescence intensity of an X-Rhod-1 dye. Shown are the mean and SEM values from at least 25 cells in each case.

We further measured the localization of the TCR/CD3 complex in subcellular membrane regions of different DNA Zipper interaction strengths. Our results indicated that immediately after adding the anti-CD3/anti-CD28 antibodies, only 39% of CD3 were localized in the membrane areas with frequent DNA Zipper interactions (Figure 6b). While at 2 min after the co-stimulation, on the same cell membranes, the majority of CD3 fluorescence (62%) came from regions with strong DNA Zipper signals. In other words, during the stimulation of TCR, CD3 may be translocated from membrane regions of moderate/weak DNA Zipper interactions to those with frequent membrane Zipper encounters, i.e., expectedly the membrane L_o phases.

Lastly, we wanted to monitor the dynamic relationship between the DNA Zipper signals and the efficiency of antibody-stimulated calcium flux during TCR signaling. Immediately after the anti-CD3/anti-CD28 antibody stimulation, we observed a rapid increase in the membrane I₆₆₀/I₅₂₀ ratiometric signal of P₇/P*, and this signal reached a plateau within ~80 s (Figure 6c). Quite interestingly, the peak value of the X-Rhod-1 fluorescence signal, i.e., the maximum cellular accumulation of Ca²⁺, was also observed at a similar time (Figure 6d). The cellular flux of Ca²⁺ lasted for another ~100 s after reaching the peak value, while the membrane I₆₆₀/I₅₂₀ signals sustained at the peak level for ~1 min and then gradually decreased to the baseline levels within ~10 min (Figure S25). Intriguingly, the spatial distribution of CD3 signals showed a similar time course to that of the DNA Zipper signals with a relatively faster increase and slower decay (Figure 6b). For the time points after 2 min of anti-CD3/anti-CD28 antibody stimulation, the percentage of CD3 signals localized within regions with weak or moderate DNA Zipper signals regions gradually increased. All these data indicated that there is clearly a strong connection between the dynamic membrane order enrichment of CD3 and the stimulation of TCR signaling.

Conclusion

Our current understanding of the plasma membrane organization and molecular interactions is still limited.^[59] The DNA Zipper probe developed here can be potentially useful in measuring dynamic interactions in live cell membranes. As an example, we chose to design a cholesterol-based DNA Zipper probe due to the critical role of cholesterols in regulating the membrane order. Indeed, our results indicated that by chemically disrupting or inducing the formation of lipid order, significant changes in the DNA Zipper signal could be observed in live cell membranes. The DNA Zipper can be applied for the imaging of membrane order in both GPMVs and living cells. These probes are thus potentially useful in studying membrane heterogeneous and dynamic patterns. While we do want to clarify that so far these images are still not super-resolved.

Emerging evidence has indicated the sophisticated roles of cholesterols in the activation of TCR signaling. Regulation of membrane order is known to modulate the efficiency of TCR activation, but contradictory results have been reported regarding the detailed roles of membrane order during this process.^[55–57,61] Our data suggested that during the anti-CD3/anti-CD28-stimulated TCR signalling in CD4⁺ Jurkat cells, the formation of membrane order occurs and leads to the generation of membrane regions with a high frequency of DNA Zipper interactions. Meanwhile, CD3 prefers to relocate into these membrane L_o phases with frequent localized DNA Zipper encounters. Interestingly, the induction of TCR-triggered calcium flux and lipid order formation are highly dynamically correlated with each other. These two processes occur almost simultaneously after the anti-CD3/anti-CD28 treatment.

We want to emphasize that a number of fluorescent probes have been developed for imaging membrane order, in most cases, based on their membrane partitioning phase preference or the solvatochromic effect (Figure S26).^[62–68] The DNA Zipper probes reported here represent a complementary approach for visualizing membrane order, i.e., via measuring membrane dynamic interactions.

In addition, there are several unique advantages of using the DNA Zipper probes for measuring membrane dynamic interactions and order. First, the DNA Zipper is compatible with standard fluorescence microscopes, costly instrumentation is not required. Meanwhile, all the DNA strands used in this study are commercially available. The probes and methods developed here can be easily adapted by researchers from different fields. Moreover, the DNA Zipper probes can spontaneously anchor onto cell membranes after a brief incubation. The whole process of incubation and imaging can be performed within ~30 min. A wide selection of donor and acceptor fluorophores with high photostability and brightness can be potentially conjugated within these modular DNA Zippers to allow more sensitive and wavelength-tunable imaging of membrane interactions and order.

The binding affinity of the DNA duplex can be precisely controlled by tuning the length and sequence of oligonucleotides. As a result, these modular and fine-tunable DNA Zippers can be potentially incorporated with a wide variety of membrane lipid complexes with different interaction strengths and stabilities. In principle, the DNA Zipper probe can be further applied in single-molecule measurements and real-time, quantitative determination of membrane dynamic interactions. Meanwhile, in addition to be modified with lipid moieties, the DNA Zipper can likely be used to study membrane protein–protein and lipid–protein interactions as well.