Shreya
Pramanik
a,
Jan
Steinkühler
ab,
Rumiana
Dimova
a,
Joachim
Spatz
c and
Reinhard
Lipowsky
*a
aMax Planck Institute of Colloids and Interfaces, 14424 Potsdam, Germany. E-mail: lipowsky@mpikg.mpg.de; Tel: +49 331 5679600
bDepartment of Biomedical Engineering, Northwestern University, Evanston, IL, USA
cMax Planck Institute for Medical Research, 69120 Heidelberg, Germany
First published on 8th August 2022
His-tagged molecules can be attached to lipid bilayers via certain anchor lipids, a method that has been widely used for the biofunctionalization of membranes and vesicles. To observe the membrane-bound molecules, it is useful to consider His-tagged molecules that are fluorescent as well. Here, we study two such molecules, green fluorescence protein (GFP) and green-fluorescent fluorescein isothiocyanate (FITC), both of which are tagged with a chain of six histidines (6H) that bind to the anchor lipids within the bilayers. The His-tag 6H is much smaller than the GFP molecule but somewhat larger than the FITC dye. The lipid bilayers form giant unilamellar vesicles (GUVs), the behavior of which can be directly observed in the optical microscope. We apply and compare three well-established preparation methods for GUVs: electroformation on platinum wire, polyvinyl alcohol (PVA) hydrogel swelling, and electroformation on indium tin oxide (ITO) glass. Microfluidics is used to expose the GUVs to a constant fluorophore concentration in the exterior solution. The brightness of membrane-bound 6H-GFP exceeds the brightness of membrane-bound 6H-FITC, in contrast to the quantum yields of the two fluorophores in solution. In fact, 6H-FITC is observed to be strongly quenched by the anchor lipids which bind the fluorophores via Ni2+ ions. For both 6H-GFP and 6H-FITC, the membrane fluorescence is measured as a function of the fluorophores' molar concentration. The theoretical analysis of these data leads to the equilibrium dissociation constants Kd = 37.5 nM for 6H-GFP and Kd = 18.5 nM for 6H-FITC. We also observe a strong pH-dependence of the membrane fluorescence.
Two broad categories of membrane proteins can be distinguished. First, integral membrane proteins have one or several transbilayer domains, which span the hydrophobic core of the bilayer membranes. Second, peripheral membrane proteins, which take part in cell signaling and membrane trafficking, can attach to one leaflet of the bilayer membranes by binding to a specific lipid or to a small cluster of several lipids.15 Likewise, to achieve biofunctionalisation of GUVs, integral membrane proteins are inserted into the membranes using detergents16 or proteoliposomes17,18 whereas peripheral membrane proteins are attached via certain anchor lipids. Here, we will consider a specific lipid anchor as provided by DGS-NTA(Ni) or NTA lipid for short,‡ which binds to poly-histidine tagged molecules via coordinate bonds. Two such molecules will be considered, His-tagged GFP and His-tagged FITC, both of which are green fluorescent, see Fig. 1.
Poly-histidine tags are generally attached to one of the terminals of the proteins for their purification.19 Hence, this specific interaction can be harnessed to attach proteins to lipid bilayers.3,6,20–24 The NTA(Ni) in the lipid head group forms an octahedral coordinate complex with poly-histidine chains. Four vertices of the octahedron are occupied by the NTA group, leaving two vertices for the binding of two imidazole side chains, see Fig. 2. The nitrogen from the imidazole side chain of histidine donates electrons for the coordinate bond. The optimal length of the histidine chain to bind to NTA(Ni) lipids consists of six residues, corresponding to the maximum of the equilibrium association constant as a function of chain length.20
![]() | ||
Fig. 2 Both 6H-FITC and 6H-GFP consist of a green fluorescent dye attached to a chain of six histidines with six side chains, corresponding to six imidazole groups. Two of the latter groups interact with the head group of the DGS-NTA(Ni) lipid forming an octahedral complex. Four vertices of this octahedron (grey) are occupied by three oxygens and one nitrogen from the head group, two vertices by two nitrogens from two imidazole side chains.20 |
The binding between His-tagged proteins and NTA anchor lipids has been frequently used in previous studies. Examples include protein crowding on vesicles,25 lipid-coated substrates for drug delivery systems,26 artificial cell adhesion,27 and high spontaneous curvature generated by the dilute concentration of surface proteins.28 In general, the Ni2+ ion can be replaced by a Co3+ ion in the NTA group. Poly-histidine chains form a stronger and less labile coordinate bond with NTA-Co3+ compared to the corresponding bond with NTA-Ni2+.29–31
GUVs can be prepared using a variety of well-established protocols such as swelling lipid layers on a substrate by hydration, electroformation,32 phase transfer across liquid–liquid interfaces,33,34 and droplet stabilized compartments;22 for a recent review, see ref. 14. All of these techniques have their advantages and disadvantages. The swelling methods involve the spreading of a lipid mixture on a substrate followed by the evaporation of the (organic) solvent. As a result, one obtains a stack of lipid bilayers which can then be swollen by hydration. Spontaneous swelling is achieved by spreading the lipid on a non-reactive surface and gently hydrating it for an appropriate period of time.35 This gentle hydration method is limited to the use of specific lipids and solution conditions.36
The influx of the hydrating solution can be enhanced by hydrogel-assisted swelling.36,37 The lipid film is deposited on the surface of the dried hydrogel. The hydrogel absorbs the hydrating solution and GUVs form rapidly at the hydrogel–aqueous solution interface. The hydrogels typically used for this purpose are agarose,37,38 polyvinyl alcohol (PVA),28,36,39 and polyacrylamide gels.40 Sometimes the hydrogel becomes incorporated into the lumen of the GUV or even within the lipid bilayer, which represents a drawback of this method because it affects the mechanical bilayer properties.38,40 To avoid these unwanted side effects, electroformation can be used to control the hydration of the lipid films by external alternating currents (ACs).32 The substrate for spreading the lipid film must be electrically conductive. Commonly used substrates include platinum wires,32,41 ITO coated glasses,42–44 stainless steel wires,45,46 copper electrodes,47 and carbon fiber microelectrodes.48
One implicit assumption that is often made in the preparation of GUVs is that these vesicles have the same composition as the lipid mixture that was initially used to grow them. This assumption should be checked, especially when the GUVs are prepared from a lipid mixture with a certain fraction of DGS-NTA(Ni). Furthermore, the binding of the poly-histidine with the NTA(Ni) lipid is also sensitive to the solution conditions. Here, we will analyze and compare different preparation techniques of GUVs in different pH conditions. As shown in Fig. 1 and 2, we will use two different fluorescent probes, 6H-GFP and 6H-FITC, where 6H stands for a linear chain of six histidines, and study the binding of these fluorophores to NTA lipids embedded in GUV membranes.
The molecular weight of the two fluorophores is quite different: GFP has a molecular weight of 27 kD whereas FITC has a much smaller molecular weight of only 389 D. Therefore, the two fluorophores are also quite different in size. As shown below, one important consequence of this size difference is that the fluorescence of FITC is strongly quenched by the NTA lipids, in contrast to the fluorescence of GFP.
![]() | ||
Fig. 3 Brightness of GUV membranes doped with DGS-NTA(Ni) for three preparation methods corresponding to the three columns: (a–c) GUV membranes exposed to an exterior solution with 320 nM of 6H-GFP. The Pt wire method leads to a strongly fluorescent membrane whereas this fluorescence is strongly reduced for the ITO glass method; and (d–f) GUV membranes exposed to an exterior solution with 240 nM of 6H-FITC. The membrane fluorescence is visible for both the Pt wire and the PVA gel methods but strongly reduced for the ITO glass method. The lipid composition used for the GUVs is POPC![]() ![]() ![]() ![]() |
In order to expose the GUVs to a constant concentration of the His-tagged molecules in the exterior solution, we control this concentration by microfluidics and a specific design of the microfluidic chip: we trap individual GUVs in the short side channels of the chip and use its main channel for the fast exchange of the exterior solution. Further details about the microfluidic approach are described in the Material and method and Fig. 10.
Two types of His-tagged molecules are studied: 6H-GFP, corresponding to a chain of six histidines attached to the N terminal of GFP,28 and 6H-FITC consisting of FITC covalently bound to the same histidine chain. The histidine chain binds to the NTA(Ni) head group of the DGS-NTA(Ni) lipid in the vesicle via a coordinate bond (Fig. 2).
As shown in Fig. 3, the observed brightness of the GUV membranes depends both on the His-tagged fluorophore and on the preparation method. For both fluorophores, the GUVs are observed to have the largest brightness when prepared by platinum wire electroformation. As will become clear further below in Fig. 6 and 7, both molar concentrations chosen in Fig. 3 – 320 nM for 6H-GFP and 240 nM for 6H-FITC – belong to the saturation regime for the platinum wire method, in which almost all NTA lipids are occupied by His-tagged molecules.
For a spherical GUV, the membrane fluorescence can be determined quantitatively by measuring the fluorescence intensity profile as a function of the radial coordinate r, which represents the distance from the center of the vesicle. One example for such a profile is shown in Fig. 4a together with the baseline intensity that interpolates smoothly between the background intensities of the interior and exterior solutions. The excess intensity of the fluorescence is equal to the difference between the total fluorescence and baseline intensities, see Fig. 4b, and the membrane fluorescence is obtained by integrating this excess intensity over the radial coordinate r. The excess intensity profiles corresponding to the GUVs in Fig. 3a–f are displayed in Fig. 3g and h.
The GUV images and line profiles show that the brightness of membrane-bound 6H-FITC is reduced compared to the one of membrane-bound 6H-GFP. In contrast, in the absence of lipids, 6H-FITC has a quantum yield that is about 1.2 larger than the quantum yield of 6H-GFP, as obtained from independent measurements, see Material and method section. This different behavior indicates that the fluorescence of 6H-FITC is quenched when this molecule is bound to the membrane or, more precisely, to an NTA anchor lipid.
![]() | ||
Fig. 6 (a) Normalised membrane fluorescence versus molar 6H-GFP concentration, as obtained by the Pt wire protocol. The fitting curve (red) is provided by eqn (6) and leads to the equilibrium dissociation constant Kd = (37.5 ± 7.5) nM for 6H-GFP. The confidence interval for Kd corresponds to the shaded strip (light grey) around the red fitting curve. The images on the right display the brightness of the GUV membranes as directly observed in the microscope for the different molar concentrations. All scale bars: 10 μm; and (b) surface coverage of GUV membranes by 6H-GFP for different preparation protocols. All GUVs were prepared with 3 mol% of NTA lipids and were exposed to a 320 nM solution of 6H-GFP at pH 7.45. The three sets of data were obtained for 19, 18, and 18 vesicles using the protocol based on Pt wire, PVA gel, and ITO glass, respectively. The solid squares represent the mean values of the coverage obtained for each data set. The numerical values for the mean coverage and the standard deviation of the data in (a and b) are given in Tables S1 and S2 (ESI†). |
![]() | ||
Fig. 7 (a) Normalised membrane fluorescence versus molar 6H-FITC concentration, as obtained by the Pt wire protocol. The fitting curve (red) is provided by eqn (6) and leads to the equilibrium dissociation constant Kd = (18.5 ± 3.7) nM for 6H-FITC. The confidence interval for Kd corresponds to the shaded strip (light grey) around the red fitting curve. The images on the right display the brightness of the GUV membranes as directly observed in the microscope for the different molar concentrations. All scale bars: 10 μm; and (b) surface coverage of GUV membranes by 6H-FITC for different preparation protocols. All GUVs were prepared with 3 mol% of NTA lipids and were exposed to a 240 nM solution of 6H-FITC at pH 7.45. The three sets of data were obtained for 17, 11, and 12 vesicles using the protocol based on Pt wire, PVA gel, and ITO glass, respectively. The solid squares represent the mean values of the coverage for each data set. The numerical values of the mean coverage and the standard deviation of the data in (a and b) are provided in Tables S3 and S4 (ESI†). |
The data in Fig. 5a and b were obtained by adding liposomes with the lipid composition POPC:
Chol
:
NTA as given by 6
:
1
:
3. Similar quenching effects were observed when the fluorophores were exposed to nickel sulphate (data not shown). On the other hand, no fluorescence quenching of 6H-FITC was observed when the liposomes contained only POPC and cholesterol but no NTA lipids, as demonstrated in Fig. S1 (ESI†).
Therefore, the fluorescence quenching of 6H-FITC is caused by NTA lipids which bind the fluorophore via their Ni2+ ion, see Fig. 2. Fluorescence quenching by Ni2+ ions has been previously reported for a variety of fluorophores.49–51 For membrane-bound 6H-GFP, such a quenching effect does not occur because the chromophore of 6H-GFP is located in the middle of the protein's β-barrel52 and well-separated from the Ni2+ ion of the NTA lipid. In contrast, the fluorescein group of membrane-bound 6H-FITC is much closer to this Ni2+ ion, see Fig. 2, which explains the strong quenching effect in Fig. 5b.
Inspection of Fig. 6a shows that the fluorescence intensity becomes saturated for large values of the molar concentration X. This saturation regime is characterized by the maximal number of 6H-GFP that can be bound to the NTA anchor lipids. For a lipid bilayer consisting of POPC and cholesterol with a molar ratio of 8:
2, the average area per lipid is 0.5 nm2 as follows from previous studies.53,54 For such a bilayer doped with 3 mol% anchor lipids, neighboring anchor lipids have an average separation of 4.1 nm. This anchor–anchor separation exceeds the lateral size of membrane-bound 6H-GFP, as explained in the Material and method. Therefore, we conclude that, in the saturation regime, the membrane-bound fluorophores have the same average separation as the anchor lipids, corresponding to an average surface coverage of 6.0 × 104 molecules per μm2, see eqn (1). In Fig. 6b, we use this saturation-based calibration to transform the fluorescence intensities into surface coverages for the three preparation methods. The scatter in the data of Fig. 6b reflects the different lipid compositions of individual GUV membranes that deviate from the overall lipid composition used for the preparation of these vesicles. The numerical values of the data in Fig. 6a and b are given in Tables S1 and S2 (ESI†).
Comparison of the equilibrium dissociation constant for 6H-FITC with the one for 6H-GFP as given by Kd = 37.5 nM reveals that 6H-FITC is more strongly bound to the GUV membranes compared to 6H-GFP. Furthermore, the different values for the saturation intensity Isat imply that the fluorescence of membrane-bound 6H-FITC is strongly reduced compared to the one of membrane-bound 6H-GFP, in agreement with the results on fluorescence quenching by NTA lipids in Fig. 5.
The GUV membranes in Fig. 7a and 6a contain the same mole fraction of 3 mol% NTA lipids. Therefore, the saturation values of the surface coverage by the two fluorophores are also identical and equal to 6.0 × 104 molecules per μm2 as obtained for 6H-GFP by the platinum wire method. In Fig. 7b, we use this saturation-based calibration to transform the fluorescence intensities into surface coverages for the three preparation methods. The scatter in the data of Fig. 7b reflects the different lipid compositions of individual GUV membranes that deviate from the overall lipid composition used for the preparation of these vesicles. The numerical values corresponding to the data in Fig. 7a and b are given in Tables S3 and S4 (ESI†).
A possible explanation for the low membrane fluorescence of GUVs prepared by ITO glass electroformation is as follows. After the deposition of lipid stock on the ITO glass surface, the lipid molecules have orientational freedom to arrange themselves into bilayers.42 The surface properties of the ITO glass are likely to affect the properties of the produced vesicles.55 The ITO film consists of indium(III) ions, which can bind to NTA.56–58 Presumably, the NTA lipid prefers to stay close to the ITO surface and forms a weak bond with indium, when the lipid stock dissolved in chloroform is deposited on the ITO surface. When the fraction of NTA is increased in the lipid stock to 30 mol%, the ITO surface becomes saturated and some excess NTA is then incorporated into the vesicle membrane.
![]() | ||
Fig. 8 Brightness of membrane fluorescence for 3 mol% NTA lipids and two types of His-tagged molecules, 6H-GFP and 6H-FITC, as observed for different pH-values of the exterior solution: GUVs exposed to 20 nM solution of 6H-GFP in (a–d) and to 120 nM of 6H-FITC in (e–h). All GUVs are prepared using the platinum wire method. All scale bars: 10 μm; and (i and j) corresponding fluorescence intensity profiles as obtained by the method described in Fig. 4. The intensity profiles in (i and j) correspond to the images in (a–d) and (e–h), respectively. In (i), the curve for pH 8.5 (blue) is masked by the one for pH 9.4 (green). |
For each pH value, the membrane fluorescence was determined via the method described in Fig. 4. The resulting fluorescence intensities of the GUV membranes are displayed in Fig. 9a and b for 6H-GFP and 6H-FITC, respectively. For 6H-GFP, the membrane fluorescence first increases and then saturates for pH values above pH 8.5. For 6H-FITC, the intensity exhibits a pronounced maximum at about pH 8.5. The numerical values of the data in Fig. 9a and b are given in Tables S5 and S6 (ESI†).
![]() | ||
Fig. 9 Membrane intensities (black squares) and background intensities (red circles) versus pH of exterior solution for GUV membranes with 3 mol% NTA lipids. All GUVs were prepared by electroformation on platinum wire: (a) For GUVs exposed to 20 nM 6H-GFP, the membrane intensity increases with increasing pH for 6.38 ≤ pH ≤ 8.5 and then saturates; and (b) for GUVs exposed to 120 nM 6H-FITC, the membrane intensity exhibits a pronounced maximum close to pH 8.5. The numerical values of the data in (a) and (b) are given in Tables S5 and S6 (ESI†). For both 6H-GFP and 6H-FITC, the background intensities (red circles) are negligible over the whole range of pH values, which demonstrates that the pH dependence of the membrane fluorescence arises from the pH dependence of the His-NTA binding affinity. |
In general, the fluorescence of a dye molecule can vary with the pH value even in the absence of lipids. In order to determine the latter pH dependence, we also measured the background intensities for 6H-GFP and 6H-FITC in the exterior solution far away from the vesicle membranes, see the corresponding data in Fig. 9. Inspection of both panels shows that the background intensity in the exterior solution is negligible over the whole range of pH values studied here. Therefore, the pH dependence of the membrane fluorescence in Fig. 9 is caused by the pH dependence of the membrane-bound fluorophores rather than by the intrinsic pH dependence of these dye molecules in solution.
One surprising result of our study is the different behavior of the two fluorophores when exposed to an increasing amount of NTA lipids. Indeed, we observed essentially no change in the fluorescence of 6H-GFP, see Fig. 5a, but a strong quenching effect on 6H-FITC as demonstrated by the data in Fig. 5b.
Both for 6H-GFP and 6H-FITC, the highest brightness of the GUV membranes was obtained when we prepared these vesicles via electroformation on platinum wire. We used this latter method to measure the dependence of the fluorescence intensities on the molar concentrations of the His-tagged fluorophores (Fig. 6a and 7a). Both sets of data are well fitted by the intensity–concentration relationship in eqn (6), which depends on only two parameters, the equilibrium dissociation constant Kd and the saturation intensity Isat.
The equilibrium dissociation constant turns out to be Kd = (37.5 ± 7.5) nM for 6H-GFP and Kd = (18.5 ± 3.7) nM for 6H-FITC which implies that 6H-FITC is more strongly bound to the GUV membranes. On the other hand, the strong quenching of 6H-FITC by the Ni2+ ions of the NTA lipids (Fig. 5b) leads to the saturation intensity Isat = 40.9 a.u. for 6H-FITC which is much smaller than Isat = 320.4 a.u. for the membrane fluorescence of 6H-GFP.
Our method to determine the equilibrium dissociation constant Kd is quite general and can be applied to the binding between other His-tagged fluorophores and/or other anchor lipids. First, one measures the fluorescence intensities as a function of the molar concentration X and then plots the normalised fluorescence intensities Iflu/Isatversus X as in Fig. 6a and 7a. Second, one fits the plotted data by the intensity–concentration relationship in eqn (6), which involves two fit parameters, the equilibrium dissociation constant Kd and the saturation intensity Isat. If the data can be well-fitted by this equation, one obtains a reliable estimate for Kd. On the other hand, if the fit turns out to be unacceptable, one can conclude that the model used to derive eqn (6) should be modified. The most important simplifying assumption of this model is that each anchor lipid can bind one fluorophore. Therefore, a poor fit will provide direct evidence that the latter assumption does not apply to the system under consideration. Possible modifications of the model include (i) crowding, i.e., the steric hindrance between the membrane-bound fluorophores, (ii) the binding of several fluorophores to one anchor lipid, and (iii) the recruitment of several anchor lipids by one fluorophore.
Another surprising outcome of our study is that it reveals a pronounced pH dependence of the membrane fluorescence for both 6H-GFP and 6H-FITC, see Fig. 9a and b. For 6H-GFP, the membrane fluorescence first increases and then saturates for pH values above pH 8.5. For 6H-FITC, on the other hand, the membrane fluorescence exhibits a pronounced maximum at about pH 8.5 (Fig. 9b). Furthermore, the observed membrane fluorescence was always much larger than the background intensity, corresponding to those fluorophores in the exterior solution that were not in contact with the lipid membranes. Therefore, the pH dependence displayed in Fig. 9 represents the pH dependence of the membrane fluorescence rather than the intrinsic pH dependence of the fluorescent molecules in solution.
Previous experiments have shown that GUV membranes exposed to nanomolar solutions of 6H-GFP acquire a large spontaneous curvature that can be fine-tuned to divide the GUVs in a controlled manner.28 As shown here, 6H-FITC has a smaller equilibrium constant and is, thus, more strongly bound to GUV membranes compared to 6H-GFP. Therefore, it will be interesting to study the spontaneous curvature generated by 6H-FITC and the shape transformations of GUVs arising from this curvature generation.
On the other hand, the His-tagged molecules reach the vesicles in the side channels only by diffusion which implies that the vesicles are shielded from the hydrodynamic flow and do not suffer any mechanical perturbations that could otherwise arise from this flow. The trapping in the side channels also helps to monitor the same GUV throughout the experiment. As shown in Movie S1 (ESI†), the complete solution exchange including the side channels takes around 3.5 min. For all experiments performed here, the pump was used to exchange 20 μL of the his-tagged molecules, corresponding to 8 times the solution volume within the microfluidic device, before the vesicles have been imaged.
![]() | (1) |
In order to estimate whether or not this saturation regime leads to crowding of the membrane-bound fluorophores, we need to compare the average separation of the NTA lipids with the lateral size of the membrane-bound fluorophores. The lateral size of membrane-bound 6H-GFP should be comparable with the diameter of its β-barrel. This diameter is about 3 nm28,61 which is smaller than the anchor–anchor separation of 4.08 nm and implies that crowding effects between membrane-bound 6H-GFP molecules can be ignored even when each NTA anchor lipid is occupied by one such molecule. The 6H-FITC peptide, on the other hand, has a linear extension of about 3.2 nm which provides an upper bound on the lateral size of membrane-bound 6H-FITC. Because this upper bound on the lateral size is again smaller than the average separation of the NTA anchor lipids, crowding effects between membrane-bound 6H-FITC molecules can again be ignored even when each NTA lipid binds one such molecule.
To measure the membrane fluorescence at different pH values, the vesicles were grown using platinum wire electroformation at pH 7.4. These vesicles were then trapped and exposed to the solution of the His-tagged molecules at different pH, ranging from 6.4 to 9.4. The pH of the solution was adjusted using HCl or NaOH. The final pH of the solutions was measured using a micro pH electrode from Mettler Toledo. The results of these measurements are displayed in Fig. 9.
dNbH/dt = κonX(Nanc − NbH) − ωoffNbH. | (2) |
![]() | (3) |
Kd ≡ ωoff/κon. | (4) |
![]() | (5) |
![]() | (6) |
For 6H-GFP and 6H-FITC, the fluorescence intensities attain the saturation values IGFP = 320.4 a.u. and IFITC = 40.9 a.u. as obtained from the HyD detector on the Leica SP8 confocal microscope. Taking into account that the data in Fig. 6a and 7a were obtained for GUV membranes with the same mole fraction of 3 mol% NTA lipids and thus with the same surface density ρanc, we obtain the ratio
![]() | (7) |
![]() | (8) |
AC | Alternating current |
DGS-NTA(Ni) or NTA for short | 1,2-Dioleoyl-sn-glycero-3-[(N-(5-amino-1-carbox-ypentyl)iminodiacetic acid)succinyl] (nickel salt) |
FITC | Fluorescein isothiocyanate |
GFP | Green fluorescent protein |
ITO | Indium tin oxide |
POPC | 1-Palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine |
PVA | Polyvinyl alcohol |
6H-GFP | Green fluorescent protein tagged by six histidines |
6H-FITC | Fluorescein isothiocyanate tagged by six histidines |
Footnotes |
† Electronic supplementary information (ESI) available: Fig. S1–S3 with control experiments and Fig. S4 with design of microfluidic chip; Tables S1–S7 with numerical values of the data displayed in the figures; caption of Movie S1. See DOI: https://doi.org/10.1039/d2sm00915c |
‡ For chemical formulas, see the list of abbreviations at the end of this article. |
This journal is © The Royal Society of Chemistry 2022 |