We present an experimental study, using the surface sensitive technique, second harmonic light scattering (SHS), to examine the influence of structure on the propensity of a molecule to passively diffuse across a phospholipid membrane. Specifically, we monitor the relative tendency of the structurally similar amphiphilic cationic dyes, malachite green (MG) and crystal violet (CV), to transport across membranes in living cells (E. coli) and biomimetic liposomes. Despite having nearly identical molecular structures, molecular weights, cationic charges, and functional groups, MG is of lower overall symmetry and consequently has a symmetry allowed permanent dipole moment, which CV does not. The two molecules showed drastically different interactions with phospholipid membranes. MG is observed to readily cross the hydrophobic interior of the bacterial cytoplasmic membrane. Conversely, CV does not. Furthermore, experiments conducted with biomimetic liposomes, constructed from the total lipid extract of E. coli and containing no proteins, show that while MG is able to diffuse across the liposome membrane, CV does not. These observations indicate that the SHS results measured with bacteria do not result from the functions of efflux pumps, but suggests that MG possesses an innate molecular property (which is absent in CV) that allows it to passively diffuse across the hydrophobic interior of a phospholipid membrane.
Following decades of study characterizing the transport of molecules across biological cell membranes, a substantial amount of knowledge has been accumulated. A detailed understanding of the functions and mechanisms of the various membrane embedded proteins (e.g., ions channels) has been revealed.1–6 Additionally, there have been numerous studies focusing on passive transport across lipid bilayers, mostly using artificial membranes in biomimetic model liposomes, also known as vesicles.7–12 Among the issues under examination are the rate of transport as well as the effects of the structure and composition of the membrane and the diffusing molecule. For instance, we have previously shown that for the same molecule of interest, the passive transport behavior can be totally different depending on the packing or rigidity of the membrane.13 Here, we present a case study of the effect of molecular structure on the propensity for passive transport across a common lipid bilayer, in which we consider membranes in both living bacteria and model liposomes.
The amphiphilic triphenyl methane dyes malachite green (MG) and crystal violet (CV) are two of the most common stains used in microbiological studies.14 CV is the well-known Gram-stain that has been used for well over a century to distinguish different classes of bacteria based upon their cellular ultrastructure.15 MG, on the other hand, is routinely used to stain the robust and poly-resistant (i.e., thermal, chemical, and photo) outer shells of dormant bacterial endospores.16 As depicted in Fig. 1, MG and CV are nearly structurally identical. They are both cations with +1 net charge. The sole difference is the third di-methyl-amino group [–N(CH2)2] in CV, which is replaced by a hydrogen in MG. This functional group substitution has a significant effect on the overall symmetry of the molecules, in which CV is of notably higher symmetry (D3h) than MG (C2v). As a result, whereas MG has a permanent dipole moment (shown as a blue arrow in Fig. 1), CV does not.
We have previously used time-resolved second-harmonic laser scattering (SHS) to interrogate the molecular interactions of MG and CV with the distinct cell wall components in living Escherichia coli.13,17–22 Despite nearly identical molecular structures, weights, and net charges, these two dyes exhibit very different interactions with cellular components. In particular, MG and CV appear to exhibit very different propensities for passively diffusing across a plasma membrane.
SHS is a nonlinear optical phenomenon by which a fraction of an intense applied field with frequency ω is scattered at 2ω after interacting with molecules whose structures do not possess a center of inversion.23–25 These molecules exhibit non-zero nonlinear polarizability, or hyperpolarizability, and are hereafter denoted SH-active. Specifically, as these SH-active molecules adsorb onto the outer leaflet of a bacterial plasma membrane, a coherent SHS signal grows in and scales in intensity as the square of the molecular surface coverage.23–25 If the molecules can traverse the membrane and adsorb onto the interior leaflet of the membrane, the scattered fields from these oppositely oriented molecular ensembles result in a coherent cancellation of the SHS signal. This characteristic time-dependent rise and fall of the SHS signal are indicative of surface adsorption and membrane transport, respectively. This interpretation has been validated in numerous studies involving both biomimetic liposomes11,26–30 and living cells,13,17–22 as well as with complementary measurements using time-resolved bright-field optical microscopy.18,19
E. coli is a Gram-negative bacterium and, as depicted in Fig. 2, can therefore be characterized as possessing two phospholipid membranes: an external outer membrane (OM) and an inner cytoplasmic membrane (CM). These two membranes are separated by a thin polymeric network of sugars and amino acids, called the peptidoglycan mesh (PM), which acts as a diffusion barrier and hinders the diffusive approach of molecules towards the CM. Additionally, the OM possesses a series of outer membrane protein (Omp) channels, which are water-filled pores that allow passive transport of small (∼600 Da) hydrophilic molecules across the OM. Conversely, except under extreme conditions (e.g., osmotic shock), direct transport across the CM requires either passive diffusion across the hydrophobic interior of the membrane or active transport across one of the various specific membrane-embedded protein channels.
Importantly, for interactions with E. coli, it has been observed that MG can adsorb onto and diffuse across the interior bacterial CM.19 Conversely, while CV likewise adsorbs onto the exterior leaflet of the CM, it appears as though it never crosses and hence never enters the bacterial cytosol.19 This interpretation was supported by complementary time-resolved bright-field optical transmission microscopy results.19 Based upon these observations, it was proposed that CV is simply unable to passively diffuse across the hydrophobic interior of a phospholipid membrane. If this is indeed the case, this suggests that, despite the numerous similarities between MG and CV (Fig. 1), there must be some distinguishing molecular characteristic which permits MG, but not CV, to passively cross a plasma membrane. Understanding such molecular properties would provide invaluable insight for the rational design and optimization of new pharmaceutical agents.
Another viable interpretation for the observed differences in the SHS results, however, could be that CV is actually able to cross the CM, but that it is rapidly kicked out of the cell before it ever has a chance to adsorb onto the interior leaflet of the membrane. As part of their survival strategies, bacteria are known to express molecular efflux pumps which help regulate the relative presence of a variety of molecular species within the cytosol by actively shuttling them out of the cell.31–33 While these pumps can be selective for overtly harmful non-endogenous species, such as antibiotics, it has been demonstrated that they can also protect the bacteria from excessive consumption of molecules abundant in the surrounding environment. For instance, Ruiz and colleagues have identified alkane efflux pumps in strains of Pseudomonas aeruginosa which allow the bacteria to thrive in liquid fuel environments.34 Specifically, while the bacteria require a portion of the alkanes from the fuel as a vital source of carbon, if left unchecked, the constant influx of alkanes would ultimately kill the bacteria. Furthermore, it is worth noting that both MG and CV fall into a class of molecules known as quaternary ammonium cations (QACs). As a family, QACs are generally known to exhibit antimicrobial properties and actually comprise the main active ingredient in the well-known disinfectant, Lysol®. Emerging results have shown the presence of QAC-specific efflux pumps which selectively remove QACs, similar to MG and CV, from Gram-positive bacteria.32 Likewise, there are analogous so-called multidrug efflux pumps in Gram-negative bacteria, such as E. coli.33 It is therefore reasonable to speculate that efflux pumps may be efficiently removing CV from the bacterial cytosol. Whether CV is unable to cross the CM, or it crosses but is rapidly removed via efflux pumps, CV will be unable to adsorb onto the interior leaflet of the CM. Consequently, the resulting time-resolved SHS signals would show similar patterns and additional experiments need to be conducted to discern the correct mechanism.
In an effort to definitively distinguish between these two possibilities, i.e., CV is either unable to cross the bacterial CM or it does so but is rapidly expelled by multidrug efflux pumps, we now examine the interaction of MG and CV with protein-free liposomes constructed from the total lipid extract of E. coli. By only using lipids (i.e., and not proteins), we ensure that there are no efflux pumps present in the system. Likewise, by using the lipids isolated from bacteria, we preserve the original molecule-lipid interaction. Aside from the necessary loss of an asymmetric distribution of lipids characteristic of membranes in living cells, liposomes composed of the total lipid extract of E. coli form a reasonable biomimetic surrogate model of a protein-free bacterial CM. Consequently, by repeating our time-resolved SHS experiments using these liposomes, we can finally determine whether or not CV is actually able to passively diffuse across a plasma membrane.
In this report, we apply time-resolved SHS to monitor the interactions between MG and CV with an ensemble of protein-free liposomes constructed from the total lipid extract of E. coli. Specifically, by considering a system specifically lacking protein, and monitoring whether or not MG and CV are able to cross the plasma membrane, it is feasible to deduce whether or not multidrug efflux pumps are responsible for prior observations19 made regarding the uptake of CV in living bacteria. The results will provide us the opportunity to examine the influence of molecular structure on the propensity for passive transport across a phospholipid membrane.
II. MATERIALS AND METHODS
A. Sample preparation
E. coli (mc4100 strain, ATCC 35695) samples were grown on Luria broth agar plates. Concentrated samples for experiments were prepared from isolated single colonies and were grown in a solution of Terrific Broth (Sigma-Aldrich) enriched with glycerin (Fisher Scientific) at 37 °C with 150 rpm continuous shaking over a duration of roughly 18 h to late-log/early stationary phase. The cells were then pelletized with gentle centrifugation (10 min at ∼1500 × g) and then washed and resuspended in phosphate buffered saline (1× PBS: 137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4) for a minimum of three times in order to remove waste products and residual grown medium. Finally, the washed pellets were resuspended in 1× PBS and diluted to an optical density of roughly OD600 = 0.2, corresponding to a cell density of ∼2 × 108 cells ml−1.
Colloidal suspensions of liposomes constructed from the total lipid extract of E. coli (Avanti Polar Lipids, Inc.), which consists of roughly 57.5% phosphatidyl ethanolamine (PE), 15.1% phosphatidyl glycerol (PG), and 9.8% cardiolipin (CL) (and 17.6% other lipid components), were generated using an extrusion method and prepared in a solution of 1× PBS. While PE is zwitterionic, PG and CL are both anionic. Consequently, the resulting liposomes exhibit a net 25% anionic surface charge. The liposome sample was characterized using dynamic light scattering. The distribution of liposomes was largely monodisperse with an effective diameter of ∼122.0 ± 10 nm. Working liposome sample densities were deduced as roughly 1013 liposomes ml−1.
Concentrated stock solutions of MG and CV were prepared by dissolution of the oxalate and chloride salt, respectively, and used as obtained from the supplier (Sigma-Aldrich). Both bacteria and liposomes were suspended in concentrated stock solutions containing 1× PBS. For each experiment, 1 ml of the sample stock was diluted to a final sample volume (i.e., containing both the dye and the cells or liposomes) of 10 ml, yielding a working concentration of 0.1× PBS (13.7 mM NaCl, 0.27 mM KCl, 1.0 mM Na2HPO4).
B. Time-resolved second harmonic laser scattering
The SHS experiment employed the 800 nm output from a Ti:sapphire laser (Coherent, Micra V, oscillator only, 76 MHz repetition rate, 150 fs pulse duration, 300 mW average power, and 4 nJ pulse energy) as the fundamental light source. Time-resolved SHS was recorded while the colloidal sample continuously flowed through a liquid flow system (i.e., in order to minimize multiple interface scattering and laser absorption loss), producing a liquid jet with a diameter of ∼1–2 mm. The liquid jet was produced by flowing the solution through a circular stainless-steel nozzle (1/16″ inner diameter). Nalgene tubing (Nalge Nunc, Inc.) was used to connect a 10 ml sample reservoir with the inlet of a motorized pump (Micropump, Inc.) and to recollect the sample in the reservoir. The colloidal sample, consisting of cells or liposomes, PBS, and dye were continuously stirred using a magnetic stirring system (Spectrocell, Inc).
The laser pulse was focused into the flowing liquid jet with a waist diameter of ∼40 µm and a Rayleigh length of 1.6 mm (∼6 nl focal volume). The scattered signal was collected roughly within the π/4 solid angle around the forward propagation direction of the fundamental beam. Effects of polarization selectivity were not explicitly examined in these experiments. The default >100:1 horizontal polarization of the laser was used as the input source, and all scattered polarizations were collected after the sample. A long-pass filter (Schott, RG695, >650 nm) was placed before the focusing lens immediately in front of the sample in order to block any higher harmonic signal produced from the preceding optics. Likewise, as both the 800 nm fundamental and 400 nm SH are scattered from the sample, a band-pass filter (Schott, BG39, 340-610 nm) and a monochromator (1 mm entrance and exit slits, 400 ± 1 nm bandwidth) were used to isolate the SH light. The SHS signal was detected with a photomultiplier (Hamamatsu, R585), pre-amplified (Stanford Research Systems SR4400), and processed using a correlated photon counting system (Stanford Research Systems, SRS SR400).
III. RESULTS AND ANALYSIS
Before examining the dye membrane interactions, it is important first to consider the composition of our measured signal. As depicted in Fig. 3, for a sample of dye and particles (i.e., cells or liposomes) dispersed in a solution of PBS, our measured signal consists of three main sources: (1) coherent second harmonic generation (SHG) scattering from the surface bound dye, (2) incoherent hyper Rayleigh scattering (HRS) from water and dye in solution, and (3) the high energy shoulder of the two-photon fluorescence (TPF) from the dye. The HRS and TPF signals persist even in the absence of particles. As a result, it is therefore feasible to isolate these background signals from the time-dependent SHG scattering signal which we will use to reveal the dye membrane interaction. Of course, this approach is only valid if the background HRS and TPF signals are time invariant sources. If either of these signals exhibit a time-dependent change, for instance, a decay due to effects of photobleaching, alternative experimental approaches would need to be implemented in order to isolate the coherent SHG scattering. Consequently, we first demonstrate that the HRS and TPF background signals are time-independent and can therefore be isolated from the time-dependent SHG scattering signal.
As shown in the right-hand side of Fig. 3, we first record the HRS from water over a duration of roughly a minute. Next, we add a concentrated aliquot of the dye and measure the HRS and TPF of the dye in solution. For both MG and CV, these signals are shown to be time-independent as they exhibit a stable plateau over a duration of 90 s. This suggests that photobleaching is not occurring to a significant extent in our experiments. This is actually quite reasonable due to the fact that the sample is a continuous flowing liquid jet (i.e., despite high peak power, we are not continuously irradiating the same volume of solution). As a result, we can isolate the time-dependent SHG signal as a perturbation to the background HRS and TPF of the bulk dye in solution. Specifically, after measuring the HRS and TPF from the dye solution, particles are then added to the sample (around t = 150 s), which gives rise to a characteristic coherent SHG signal which can be analyzed to deduce specifics of the dye membrane interaction.
As described previously,19 Fig. 4 depicts the representative time-resolved SHS signal collected following the interaction of MG (green trace) and CV (violet trace) with a colloidal suspension of living E. coli. The first 10 s corresponds to the incoherent HRS and TPF background signal from dye molecules randomly dispersed in bulk solution. For both MG and CV, the addition of E. coli at ∼t = 10 s resulted in an immediate spike of the SHS signal as the dye instantly saturated the exposed outer leaflet of the bacterial OM. This was subsequently followed by a rapid signal decay which corresponds to the dye crossing the bacterial Omp channels and adsorbing onto the interior leaflet of the OM. The SHS signals are nearly identical until about t = 25 s. Beyond that, the measured responses begin to deviate significantly.
For MG, starting around t = 25 s, a second rise in the SHS signal is observed to grow which corresponds to adsorption of MG onto the outer leaflet of the CM. This continues until roughly t = 100 s, after which the signal begins to slowly decay, over a period of several hundred seconds, as MG passively diffuses across the CM and gradually adsorbs onto the interior leaflet of the CM. Conversely, the measured SHS from CV continues to decay for twice as long until roughly t = 50 s. From there, similar to the MG response, a second rise in the SHS signal slowly begins to rise in until about t = 400 s. Consistent with the MG response, this secondary rise of the signal can be assigned to a slow adsorption of CV onto the exterior leaflet of the CM. The fact that the CV signal requires an additional 300 s to saturate the exterior of the CM suggests that CV exhibits a comparatively strong interaction with the bacterial PM (i.e., it takes a significantly longer time for CV to arrive at the CM). Afterwards, the measured SHS signal is observed to remain effectively constant, even beyond t = 2000 s, suggesting that CV is either unable to diffuse across the CM or it is rapidly removed from the cytosol after crossing the CM. In either case, CV does not adsorb onto the interior surface of the CM leaflet. Consequently, the SHS signal from CV on the exterior leaflet of the CM is observed to persist indefinitely.
In order to discern whether or not CV is able to passively diffuse across a phospholipid membrane, we now consider the SHS responses of MG and CV, respectively, in a protein-free biomimetic model system. Specifically, Fig. 5 depicts representative time-resolved SHS signals measured following the interaction of MG (green trace) and CV (violet trace) with a colloidal suspension of liposomes constructed from the total lipid extract of E. coli. Similar to the bacterial results (Fig. 4), the liposome experiments began by first measuring the incoherent HRS and TPF background signal from MG or CV in the bulk solution only. Likewise, following addition of liposomes into the sample reservoir (around t = 40 s), the measured SHS signal exhibited an immediate spike in intensity as the exposed exterior surface of the liposomes was instantaneously saturated with dye molecules. However, all similarities between the MG and CV responses end here. Following instantaneous saturation, the SHS signal from MG (green trace) begins to rapidly decay as MG quickly diffuses across the membrane and starts to adsorb onto the interior leaflet of the liposome. Fit analysis using a single exponential function yields a decay time constant of τ = 80.3 ± 4.9 s. A similar analysis of the signal decay corresponding to transport across the bacterial CM (Fig. 4) yields a slightly more relaxed decay time constant of τ = 203.1 ± 10.4 s, suggesting that the liposome membrane is more permeable than the bacterial membrane. Conversely, similar to the bacterial CM results, following instantaneous saturation, the SHS signal from CV (violet trace) simply maintains a constant intensity. Given that we do not measure the signal plateau for an infinitely long duration, it is feasible that CV may simply cross the membrane at an extremely slow rate, resulting in an equally slow signal decay. As a rough estimate of this possibility, we apply a similar fit analysis of the possible decay within the noise of the measured signal. Such an analysis yields a massive decay time constant of τ = 5.4 × 109 s, with an even larger uncertainty (5.7 × 1014). Put another way, CV has a decay time constant of at least 158 years. Given the complete absence of proteins in the liposome samples and, in particular, efflux pumps, the persistence of the CV SHS signal suggests that CV is not able to passively diffuse across the liposome membrane.
Before we examine whether or not CV is able to transport across a membrane, it is perhaps fruitful to first discuss the likely mechanism by which these dyes are crossing. In recent years, there has been considerable attention to the phenomenon of cell-penetrating peptides (CPP’s).35,36 Briefly, CPP’s are typically short poly-cationic peptides which exhibit favorable electrostatic binding with the anionic surface of cell membranes. Above a critical concentration, the peptides can begin to aggregate and form transient aqueous channels (or pores) across the hydrophobic interior of the membrane.35,36 It is worthwhile considering whether or not our triphenylmethane dyes could be utilizing a similar mechanism to cross the membranes under study. While MG and CV are considerably smaller than typical CPP’s, it is worth noting that our lab has recently reported comparable behavior in the similarly sized antibiotic cation, azithromycin (MW = 748.996 Da), interacting with the membranes in E. coli.22 Specifically, it was observed that the transport rate of MG crossing the bacterial CM increased by an order of magnitude following prolonged administration of azithromycin above the minimum inhibitory concentration.22 It was suggested that azithromycin may be employing a mechanism analogous to the so-called carpet disruption model, in which intercalation of polycationic compounds between the hydrophilic head groups disrupts the packing of the phospholipids and leads to local thinning of the membrane. The crucial point here is that a mechanism involving the formation of a transient pore should be expected to result in rapid transport. The fact that transport across the bacterial CM (as well as the membrane in liposomes) is observed to be two orders of magnitude slower than that observed for the water-filled Omp channels in the bacterial OM suggests that there is no fast route across the bacterial CM. Moreover, it has previously been observed that transport of MG across liposome membranes is sensitive to the concentration and identity of the counterion.37 Briefly, it was observed that small atomic counterions (such as chloride and bromide) reduce the transport rate of triphenylmethanes (such as MG and CV) by forming a loose ion pair which stabilizes the MG cation in bulk solution. Given that transient pore mediated transport permits the solute molecules to remain in the aqueous phase, such mechanisms should be unaffected by such counterion effects. Consequently, it is reasonable to conclude that the transport mechanism for our dyes is direct diffusion across the hydrophobic interior of the membrane.
A. CV does not diffuse across plasma membranes
In a prior study monitoring the interactions between CV and living bacteria, evidence was presented suggesting that CV is unable to passively cross a plasma membrane.19 Specifically, as depicted in Fig. 4, following adsorption onto the exterior leaflet of the bacterial CM, a time-dependent increase in the SHS signal was observed and continued to rise until the bacterial membrane surface was completely saturated with CV. Following the onset of surface saturation, the measured SHS signal exhibited a plateau in time and remained constant thereafter. If CV cannot cross the CM, it will have no access to the interior leaflet of the CM and will therefore be unable to adsorb onto this surface. Consequently, the measured SHS signal from CV adsorbed onto the exterior surface of the CM would persist indefinitely. However, as depicted in Fig. 2, an alternative explanation could be that CV may indeed be able to cross the CM, but that it is rapidly expelled from the cytosol via one of the known bacterial efflux pumps. Provided that the rate of efflux (kefflux) is sufficiently faster than the rate of surface adsorption (kads), there would never be an appreciable concentration of CV adsorbed onto the interior membrane surface; hence, there would be nothing to cancel the SHS signal from CV adsorbed on the exterior membrane surface. Overall, the only conclusion that can be drawn with any certainty is that CV does not adsorb onto the interior leaflet of the CM.
We first consider the possibility that CV can indeed cross the CM, but that it is rapidly removed from the cytosol via efflux pumps. Based upon the bacterial results shown in Fig. 4, given that MG is observed to cross the CM and saturate the interior leaflet of the CM, this would indicate that the efflux pumps are selective for CV but not MG. This is not an unreasonable possibility. Given that efflux pumps are an active transport system, this suggests that the molecular target likely binds with a recognition component of the efflux pump machinery. While MG and CV are structurally similar, as shown in Fig. 1, they are obviously not identical. In fact, CV has higher structural symmetry than MG (i.e., D3h vs C2v) due to the presence of the third di-methyl-amino group [–N(CH3)2]. It is therefore possible that CV could exhibit preferential binding with an efflux pump and hence be selectively ejected from the cell.
This idea can be tested based on the observation on the interactions of MG and CV with protein-free liposomes, constructed from the total lipid extract of E. coli. Given that these liposomes contain the same lipids and lipid ratios present in the bacterial study (Fig. 4), they represent ideal biomimetic surrogates of the bacterial CM in which all protein (including the contentious efflux pumps) have been completely removed. In this way, we can definitively test whether or not the CV interactions observed in Fig. 4 stem from the normal activity of efflux pumps. As depicted in Fig. 5, similar to the bacterial CM results, MG is observed to adsorb onto and readily transport across the liposome membrane. This is a reasonable and fully expected result. After all, MG is able to cross the CM in living bacteria, and living cells are known to possess more rigid membranes than liposomes, due in part to the presence of embedded proteins and rigidfying molecules, such as cholesterol or the bacterial equivalent, hopanoids.27,38,39 More crucially, it is observed that while CV readily adsorbs onto the exterior surface of the liposome membrane, as evidenced by the immediate rise in the SHS signal, it once again appears as though CV is unable to cross the membrane (i.e., the time-dependent signal does not decay). Given that the liposomes do not possess efflux channels, or protein of any kind, the absence of a transport signal can no longer be attributed to the rapid removal of CV from the liposome interior. CV is simply unable to passively diffuse across the hydrophobic interior of the membrane.
While it is now evident that efflux pumps cannot account for the observed CV interaction with the bacterial CM, there is yet a third possibility that we need to consider. Specifically, the asymmetry of the phospholipid bilayer could prevent CV from adsorbing onto the interior leaflet of the bacterial CM. That is, if the interior leaflet of the CM is sufficiently different from the exterior leaflet, it is possible that CV may simply exhibit unfavorable binding to that surface. While it is true that membranes in living cells possess asymmetric phospholipid membranes (i.e., the lipids on the exterior leaflet are known to be different than the lipids on the interior leaflet),40 this is unlikely to cause drastically different adsorption behaviors. In particular, for both MG and CV, the dominant surface adsorption interaction is electrostatic in nature.41 Cationic MG (or CV) is attracted to the anionic functional groups of the phospholipid heads. The fact that we observe comparable adsorption of MG to both leaflets of the bacterial CM strongly suggests that both surfaces exhibit a net anionic charge. Therefore, while the lipid composition of the two leaflets of the bacterial CM is known to be different, the fact that MG is able to electrostatically bind to both surfaces suggests that cationic CV should be able to do so likewise.
Moreover, the liposome results depicted in Fig. 5 permit a more refined experimental test of this hypothesis. Specifically, while the extrusion method used to prepare the liposomes in this study is able to construct a unilamellar membrane bilayer, the process largely homogenizes the lipid composition of the resulting membrane. Transmembrane asymmetry is actually regulated by a combination of passive and active mechanisms.42,43 Whereas asymmetry dictated by physical properties (e.g., hydrogen bonding and charge distribution) will certainly persist, all protein mediated active mechanisms would no longer be available.44 Consequently, the lipid distribution in the resulting liposomes should be significantly more homogenized compared to the membrane in the living bacterial cell. As distinct from the bacterial CM, the lipid composition of the membrane leaflets in our liposomes should be very similar. Consequently, the fact that CV is clearly shown to adsorb onto the exposed exterior surface of the liposomes (i.e., the measured SHS signal rises) implies that it would certainly be able to adsorb onto the interior leaflet. As a result, the absence of a decay in the time-resolved SHS signal for the liposome system therefore definitively indicates that CV is not able to cross the membrane (Fig. 6). Furthermore, as noted above, given that membranes in liposomes are known to be less rigid than those in living cells,27,38,39 the fact that CV is unable to diffuse across the comparatively fluid membrane of the liposome strongly suggests that CV should likewise not be able to diffuse across the much more rigid membrane found in a living cell.
One further important distinction which we have not yet discussed is that, despite their strong similarity, MG and CV exhibit very different pH dependency. Under sufficiently basic conditions, both MG and CV react with hydroxyl ions to form neutral species called carbinols, in which the hydroxyl attacks the central carbon of the triphenylmethane.45 The pKa’s for these reactions are well separated at 6.9 and 9.4 for MG and CV, respectively.45 Consequently, under the experimental conditions of the current study (pH = 7.3), only ∼30% of the MG population would exist in the cationic form. Conversely, 100% of the CV population would be cationic. It is tempting to speculate that this difference (i.e., MG simultaneously exists in a cationic and a neutral form) could reasonably account for the different tendency to traverse a membrane. That is, whereas the hydrophilic cation may struggle to cross the membrane, the hydrophobic carbinol should cross with relative ease. Then, once the carbinol was in the interior domain of the cell or liposome, it could quickly re-establish acid-base equilibria and produce a population of the cation, which could then bind to the interior surface of the membrane and produce the observed SHS signal. Given that CV is purely cationic under these conditions, this proposed mechanism suggests that it should not be able to cross the membrane. In principle, this hypothesis can be tested with pH-dependent experiments. Focusing on MG, running the experiment in an acidic solution (pH = 4) would convert all of the neutral carbinol to the cationic form. Under these conditions, in which there is no neutral MG present, MG should no longer be able to cross the membrane (similar to CV under neutral pH). It is of interest to note that numerous such experimental data are actually already available in the existing literature. For instance, much of the pioneering work out of the Eisenthal Lab studied the transport of cationic MG across liposome membranes under acidic conditions (pH = 4).26,27,37,41,46 In all those cases, cationic MG is definitively shown to cross the liposome membrane. As a result, while acid-base reactivity would be a satisfying explanation, it cannot account for the different behavior observed here.
In addition to the transporting ion, Eisenthal’s Lab has previously shown that counter ions significantly affect the efficiency of membrane transport.37 This is especially pertinent to the current study in which our MG and CV samples had different counterions (i.e., oxalate and chloride for MG and CV, respectively). It was previously shown that small inorganic counter ions, such as chloride, resulted in a comparatively reduced MG transport rate.37 It was suggested that, as the inorganic ions were sufficiently small to fit within the binding pocket of MG’s central charged carbon, a loose ion pair could be formed which stabilized MG in solution (hence reducing the propensity for membrane transport).37 It is important to note, however, that in addition to the dye counter ions, all of our experiments were run in a phosphate buffer solution (i.e., a final concentration of roughly 0.1× PBS). Consequently, in addition to the micro-molar ion concentrations of oxalate (MG experiments) and chloride (CV experiments), all experiments contained a comparatively massive milli-molar concentration of chloride ions (i.e., yielding a buffer to dye counter ion ratio of roughly 1400:1). In this respect, it is reasonable to suggest that all experiments in the current study had the same counter ion (chloride). While this should certainly reduce the deduced transport rates, there is no indication that this would prevent transport altogether. Indeed, the study from Eisenthal and colleagues demonstrated a counter ion concentration dependence, in which the MG transport rate decreased as the concentration of chloride increased.37 However, even for counter ion concentrations of 100 mM (i.e., 10 times larger than the concentration used in the current study), MG still readily crossed the membrane.
B. Effect of molecular structure on diffusion across a membrane
Taken together, the liposome (Fig. 5) and bacterial (Fig. 4) results clearly demonstrate that CV is physically unable to passively diffuse across a phospholipid membrane, regardless of whether the membrane is in a living cell or in a purely synthetic liposome. Parenthetically, this also demonstrates why CV (i.e., the Gram-stain) can be used to differentiate bacteria as either Gram-positive or Gram-negative.
Given the structural similarity between MG and CV, the interesting observation is the fact that MG, but not CV, is readily able to passively diffuse across a phospholipid membrane. This is especially of interest from a pharmaceutical perspective in which it is often necessary to design non-endogenous molecules capable of passively entering cells. The question now is, what is the structural characteristic contained within MG, but absent from CV, that permits passive diffusion across a plasma membrane? Due to the three hydrophobic phenyl rings located at the periphery of both MG and CV, it seems reasonable that both molecules, based on functional groups, should behave similarly. Furthermore, given that MG and CV are of similar size and mass, they have identical cationic charges (+1), and nearly identical structure, there are few possibilities that could reasonably account for this behavior. As shown in Fig. 1, we note that CV has higher molecular symmetry than MG (i.e., D3h vs. C2v, respectively). Based on this, one important distinction is that MG has a permanent dipole moment, whereas CV does not. While we have now definitively shown that CV is unable to cross a plasma membrane, it is reasonable to speculate that molecules not possessing a permanent dipole moment may not be able to passively diffuse across a membrane. This propensity will need to be further tested before it can be suggested as a general rule. Nevertheless, to the best of our knowledge, there are not yet any experimental examples of a molecule lacking a permanent dipole passively diffusing across a membrane. A more fundamental question is why this may even be the case, that is, in what way does would a permanent dipole help a molecule traverse the hydrophobic interior of a membrane? Additionally, if a permanent dipole is required to cross a membrane, can an induced dipole initiate a similar effect? It is our hope that our preliminary experimental observations will inspire a complementary theoretical analysis of this phenomenon.
In summary, we have applied time-resolved SHS to examine the relative propensities for two molecules with similar size, molecular weight, charge, and chemical functional groups, but with different symmetries and polarities to passively diffuse across a phospholipid membrane. Experiments with living bacteria revealed that while MG was able to transport across the cytoplasmic membrane of E. coli, CV was not. Furthermore, in an effort to determine if these observations were obscured by the activity of bacterial efflux pumps, the SHS experiments were repeated with protein-free biomimetic liposomes constructed from the total lipid extract of E. coli. Similar to the bacterial results, it was observed that while MG was able to cross the liposome membrane, CV was not. Taken together, these results suggest that CV is physically unable to passively diffuse across a phospholipid membrane. While additional investigation is necessary, it is speculated that the symmetry allowed permanent dipole of MG (which is notably absent in CV) may be responsible for MG’s innate ability to passively diffuse across a phospholipid membrane.
This work was supported by the National Science Foundation (Grant No. CHE-1465096).