Introduction

Synthetic transmembrane anion transporters are an area of intense current interest due to their potential application in the treatment of diseases1,2,3. The potential for anti-cancer activity4, and to replace the anion transport process in patients with genetically impaired ion channels5, has led to a significant rise in the number of small molecule anion transporters developed, such as indole-based receptors6, the ortho-phenylenediamine-based bisureas7, as well as many others8,9,10,11,12. However, the natural product prodigiosin (Fig. 1) remains a benchmark for new anion transporters due to its potent anion transport properties in lipid bilayer membranes13,14,15,16. Prodigiosin has been shown to bind H+/Cl− through a combination of hydrogen bonds and electrostatic interactions17, allowing it to facilitate H+/Cl− symport through membranes resulting in pH dissipation in both model liposomal and cell assays18,19,20,21. In cells, the dysregulation of ion homeostasis through chloride influx, which can be facilitated by anion transporters22, can induce cell shrinkage and leads to the apoptotic cell death pathway23,24,25,26. Another consequence of the dysregulation of intracellular chloride concentrations is disruption of autophagy, through the interruption of lysosomal enzyme activity caused by altering the lysosomal pH27. The disruption of autophagy by small molecule anion transporters results in the suppression of cell proliferation which is more pronounced in tumour cells28. A second family of recognised H+/Cl− cotransporters, the tambjamines29,30,31, have also been shown to induce lysosomal dysfunction29 and autophagy blockage32 as well as cancer cell death3.

Figure 1
figure 1

Schematic of prodigiosin (left) and perenosin (right) showing the chemical structures of these H+/Cl− cotransporters as free receptors and their respective HCl complexes. Three modification positions for the perenosins reported in this work are highlighted R1, R2 and R3.

Full size image

The encouraging antitumor activity displayed by members of the prodiginines family33 has promoted the development of new synthetic derivatives such as obatoclax34. Recently, we have developed a new class of anion transporters, the “perenosins”, with structures inspired by that of prodigiosin (Fig. 1)35. The incorporation of an imine linker in the structural design of perenosins is beneficial for further pharmacokinetic studies, especially the metabolic breakdown pathway after administration.

Perenosins consist of a dimethylpyrrole unit linked through an imine moiety to an indole, benzimidazole or indazole heterocyclic functional group, which are the basis of other anion receptor systems36, 37. Perenosins function as H+/Cl− cotransporters in a similar fashion to prodigiosin, as protonation of the imine nitrogen leads to binding of the chloride anion. Liposome-based anion transport studies showed that the indole derivatives such as compound 1d (Fig. 2) performed as one of the more effective transporters, with superior cytotoxic properties among the initial compounds studied.

Figure 2
figure 2

Structures of the perenosins reported in this work; 1a–e from the first generation are used for comparison, 2–6 are new analogues based on the same indole-pyrrole scaffold of 1, and 7–8 are based on a modified bis-indole scaffold.

Full size image

Encouraged by the outstanding transmembrane transport and cytotoxic properties of compound 1d, we have developed a new family of perenosins 2–6 based on the indole-pyrrole scaffold of 1a–e, and bis-indole perenosins 7–8. Here, we report detailed studies of the chloride anion binding, pK a in aqueous and lipid bilayer environments and transmembrane transport capabilities (and mechanisms) of the indole-based perenosins. We have also performed in-depth cell studies of the cytotoxic properties of these compounds (1a–e and 2–8), and have comprehensively elucidated the transport mechanism in liposomes as well as the cell death pathway induced by perenosins for the first time.

Results and Discussion

Indole-Based Perenosins

Design and synthesis

The pyrrole-indole derivatives of the first generation of perenosins featured modifications at one site, R1 (Fig. 1). Adapting the reported stepwise synthetic approach allowed incorporation of a variety of substituents at the R2 and R3 positions, as highlighted in Fig. 1 (see Supplementary Information for full synthetic schemes). Briefly, non-commercial indoles were synthesised from a 2-nitroaniline starting material (with the respective R1 substituent) via iodination, followed by a Sonogashira coupling with an alkyne (corresponding to the R2 substituent), and finally a base assisted cyclisation step. Manipulation of the pyrrole through the Knorr pyrrole synthesis afforded the pentyl-substituted moiety at the R3 position. Hydrogenation of the nitroindole and subsequent condensation with the pyrrole aldehyde gave the final products in good yields.

The new family of perenosins reported here expands the library of perenosin derivatives. Introducing alkyl chains of different lengths at R2 for compounds 2 (pentyl) and 4 (propyl) allowed investigation into the effects of encapsulation of the binding site on the transport properties of the perenosins. Another advantage of using the R2 position for alkylation was that other substituents, such as a CF3 group (compound 3) which has been shown to enhance transport capabilities38, 39, could be positioned at R1 whilst retaining the alkyl chain in the structure. R3 was alkylated allowing investigation into the effects of optimal lipophilicity40,41,42, showing the differences between having a pentyl chain on the pyrrole rather than indole (compound 5) or on both (compound 6). As indole-NH groups are more acidic and hence are better hydrogen-bond donors than the pyrrole-NH groups43, 44, we have also developed two bis-indole derivatives (7 and 8) as an alternative perenosin scaffold to potentially improve anion binding and its effect on transmembrane transport.

Previously, we reported the X-ray crystal structure of the HCl complex of the unsubstituted analogue (R1 = H) of 1a 35 (Fig. 3a), here we present the X-ray structure of free perenosin 2 (Fig. 3b), see the Supplementary Information for the single crystal X-ray structures of compounds 1d and 4. The structure of compound 2 reveals the formation of an orthogonal ‘narcissistic’ dimer45,46,47 in the solid state via intermolecular hydrogen bonds between the imine nitrogen acceptor and two NH donors from pyrrole and indole.

Figure 3
figure 3

Single-crystal X-ray structures of (a) 1a·HCl complex (CCDC: 1441636)35 with N1···Cl, N2···Cl and N3···Cl distances displayed as red dashed lines with measurements (Å) in black; and (b) 2 with N1···N2′, N3···N2′, N1′···N2 and N3′···N2 distances displayed as red dashed lines, with measurements (Å) in black.

Full size image

Anion Binding Studies

Proton NMR titrations were performed in DMSO-d 6/0.5% H2O, to assess the ability of compounds 1d and 2–8 to bind chloride, added as the tetrabutylammonium (TBA) salt, in solution. Free perenosins showed weak binding of the chloride anion. However, in the presence of one equivalent of HPF6, strong binding to chloride was observed through downfield shifts of the pyrrole-NH, indole-NH and the protonated imine-NH+. All three resonances were fitted globally to the 1:1 or 2:1 (sandwich) binding models using Bindfit48. Apart from 1d; the data of perenosins 2–6 are more complex and did not fit well to the 1:1 model, however, it can be fitted to the 2:1 binding model (see Supplementary Information) with an improvement in the quality of fit (covfit) by >6 times (see Table 1). The curve fitting analysis demonstrated that a 2:1 sandwich-binding mode for perenosins 1d and 2–6 is plausible, with the protonated receptors showing a 2:1 binding mode in the presence of small amounts of chloride, and a 1:1 complex at higher chloride concentrations. The 1:1 and 2:1 binding modes are supported by the PM6 optimised structures of perenosin 2, demonstrating an octahedral 2:1 host:guest complex (Fig. 4).

Table 1 Overview of association constants for the complexation between protonated perenosins (1d and 2–8) and Cl− (as TBA salt) in DMSO-d 6/0.5% H2O at 298 K.
Full size table
Figure 4
figure 4

Theoretical (PM651) optimised structures performed using Gaussian0952 for compound 2 in both the 1:1 and 2:1 binding mode; shown in space-filling models.

Full size image

By comparing the derived association constants from the 2:1 fitting analysis for perenosins 1d and 2–6, 2 emerged to be the best receptor for binding to chloride, with the largest K 11 (and K 21 for most cases but smaller than 3, to be discussed later), as well as the largest overall binding constant β21. The superior chloride binding of 2 suggests that the encapsulating effect of the binding site by pentyl chain at the R2 position can enhance the binding affinity for chloride. This hypothesis is further proven by the decrease in binding for 4 (with a shorter propyl chain, hence less encapsulating) and more significant decrease for 5 (absence of an alkyl chain at the R2 position); whilst compound 6 exhibits improved binding compared to 5 when the pentyl chain is re-introduced at the R2 position. Interestingly, only compound 3 exhibits strong positive cooperativity49, where K 21 is much greater than K 11 50, hence a preferential 2:1 (perenosin:chloride) complex formation occurs (see speciation ratio plots in Supplementary Information).

Addition of one mole equivalent of HPF6 to bis-indole perenosins 7 and 8 resulted in partial protonation of the imine nitrogen, approximately 20% and 10% respectively. Subsequent titration of 7 and 8 in the presence of HPF6 (1 mole equiv.) with tetrabutylammonium chloride induced a downfield shift of only one indole-NH (NH1 or NH1′ shown in Fig. 3), and resulted in weaker binding to chloride compared to the indole-pyrrole perenosins. Although the titration isotherms of 7 and 8 can be fitted to both 1:1 and 2:1 binding modes, the derived association constants must be viewed with caution.

It must be noted that the formation of 2:1 and 1:1 host:guest complexes during the titration binding studies resulted in a degree of uncertainties in the derived binding constants due to the complexity of the data. We acknowledge that more in-depth studies (such as solvent effects and protonation equilibrium in the same solvent environment) are required to gain better understanding of the 2:1 sandwich binding and the encapsulation effects. However, the focus of this work is on the transport and biological activities, hence it is more suitable to further explore the binding properties in a separate work.

pK a Studies

The pK a at which protonation of the imine (N2, see Fig. 3) occurs gives an indication of the optimum pH for the binding of chloride anions. This is an important consideration in anti-cancer drug development as the external pH environment of cancer cells is mildly acidic (pH 6.7)53 compared to that of healthy cells (pH 7.2–7.5)54. This can be exploited to selectively target cancer cells and potentially reduce the cytotoxic effects of perenosins when used as therapeutics for anti-cancer treatment. The apparent pK a values, shown in Table 1, for compounds 1d and 2–8 were determined both in aqueous solution (pK a (aq)) and in the presence of a 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) lipid bilayer membrane (pK a (mem)). UV spectrophotometric titrations were conducted by varying the pH condition from neutral to acidic with the titration of either HPF6 or HCl. While the neutral perenosin has an absorbance maxima of ~380 nm for 1d and 2–6 and ~330 nm for 7–8, protonation resulted in a red shift to ~420 nm and ~400 nm respectively.

The apparent pK a values were derived from nonlinear curve fitting of the absorbance plotted against the pH; global fitting of both wavelengths corresponding to the neutral and protonated species were performed whenever feasible, see Supplementary Information for details. The pK a values obtained for 1d and 2–8 are lower in the presence of a lipid bilayer, indicating that the lipid bilayer stabilises the neutral form of the compound to some extent55, 56.

Transmembrane Transport Studies

Cl−/NO3 − antiport at pH 7.2

The chloride transport properties of the new perenosins 2–8 were compared to 1d using the chloride/nitrate exchange assay monitored by a chloride ion selective electrode (ISE). Typically, unilamellar vesicles were prepared from POPC, containing an intravesicular NaCl solution (489 mM with 5 mM phosphate buffer at pH 7.2) suspended in a NaNO3 solution (489 mM with 5 mM phosphate buffer at pH 7.2). All perenosins were added as a DMSO solution and the rate of chloride efflux was monitored by the chloride ISE. At the end of the experiment (300 s), detergent was added to lyse the vesicles and obtain the 100% chloride efflux reading. Dose response studies of 1d and 2–8 were performed to allow Hill analysis, giving the EC50 values and Hill coefficients as shown in Table 2.

Table 2 EC50 values shown for the Cl−/NO3 − antiport assay, the NMDG-Cl assay alone, with gramicidin (Gra) and with oleic acid (OA), with the chloride selectivity (SG) and the factor of enhancement (FOA) included.
Full size table

Perenosin 1d, bearing a pentyl group at R1 position, achieved moderate transport efficiency with an EC50 of 0.122 mol%. On the other hand compound 2, with pentyl substitution at R2 position, has a remarkable EC50 value of 0.0043 mol% being the most effective transporter of the entire series, and is comparable to the natural product and synthetic analogues of tambjamines39. However, introduction of the pentyl group at the R3 position in compound 5 decreased the transport capabilities as compared to compound 2 (but the compound still performed better than 1d), with an EC50 value of 0.0227 mol%. While perenosins 1d, 2 and 5 all possess a pentyl chain at R1, R2 or R3 positions with very similar log P values (4.9–5.3), the respective transport efficacies are significantly different. This outcome highlights the beneficial effect of strategically positioning the pentyl chain to improve the mobility within the lipid bilayer and more importantly, enhancing the encapsulation of the binding site. Compound 3 achieved a maximum chloride efflux of 35% (after 300 s) at a high loading of 10 mol%, so an EC50 value could not be determined; this is likely due to poor solubility in aqueous solution as was observed in the pK a studies. Compound 4 bearing a shorter propyl group at R2 position resulted in a slightly poorer EC50 of 0.0111 mol% c.f. 2, which can be attributed to the lower lipophilicity and lesser encapsulating effect by the shorter alkyl chain. The most lipophilic compound 6, with two pentyl groups at the R2 and R3 positions, resulted in a decrease of transport capabilities with an EC50 of 0.0852 mol% and this is presumably due to the compound’s unfavourably high lipophilicity and poor solubility. Finally, compounds 7–8 gave weak transport activity at pH 7.2 and Hill analysis could not be used to calculate EC50, potentially due to a number of factors; (1) the pK a values being two pH units lower than the pH condition of the experiment for compound 7, (2) weak chloride binding and (3) unfavourably high lipophilicity. Overall, the new perenosins bearing the indole-pyrrole scaffold achieved better transport efficiencies than 1d, the best perenosin transporter previously reported.

pH dependent Cl−/NO3 − antiport

The external pH of cancer cells (pH 6.753) is slightly acidic compared to healthy cells (pH 7.2–7.554). Therefore, for the perenosins to show a degree of cytotoxic selectivity towards cancer cells over healthy cells it is desirable for cellular HCl cotransport to be more effective in the mildly acidic cancerous cell environment. To investigate the effect of pH on the chloride transport capabilities unilamellar vesicles were loaded with NaCl (489 mM) and suspended in NaNO3 (489 mM), and the internal and external solutions were buffered to pH 4.5, 6.5, 7.2 and 10 using various buffers (more detail can be found in the Supplementary Information). Compounds 1d and 2–8 were then tested in a single point screen at 0.1 mol% loading (Fig. 5). At pH 4.5 the perenosins exist predominantly as the active protonated species and so all compounds show moderate to high chloride transport properties, whereas, there is no chloride transport observed at pH 10 (see Supplementary Information for results). For all compounds, except 7, an increase in chloride transport is observed at pH 6.5 compared to pH 7.2, suggesting that these perenosins could potentially be more selective toward cancer cells (Fig. 6).

Figure 5
figure 5

Chloride efflux achieved from compounds 1d and 2–8 at 0.1 mol% loading from unilamellar vesicles loaded with NaCl (489 mM) suspended in NaNO3 (489 mM) buffered to pH 6.5 (left) and pH 7.2 (right) with sodium phosphate salts (5 mM). All data are the average of three repeats, performed using the same batch of vesicles, with error bars corresponding to the SD.

Full size image
Figure 6
figure 6

Schematic of the cationophore coupled KCl efflux assays measured by chloride ISE, from unilamellar POPC vesicles loaded with KCl (300 mM) and suspended in K2SO4 (150 mM), buffered to pH 7.2 with potassium phosphate salts (5 mM). (a) Overall KCl efflux induced from a combination of electrogenic K+ transport facilitated by valinomycin (Vln) (0.1 mol%) and electrogenic Cl− transport facilitated by the transporter (T). (b) Overall KCl efflux induced from electroneutral transport processes; K+/H+ antiport by monensin (Mon) (0.1 mol%), and Cl−/H+ symport and Cl−/H+ antiport facilitated by transporter (T). (c) Chloride efflux facilitated by compound 2 (0.1 mol%). (d) Chloride efflux facilitated by prodigiosin (0.01 mol%). All data are the average of three repeats, performed using the same batch of vesicles, with error bars corresponding to the SD.

Full size image

H+/Cl− symport mechanism studies

The previously reported perenosins 1a–e were shown to function in the same way as prodigiosin and facilitate chloride transport when protonated through H+/Cl− symport (see Fig. 1), as demonstrated using a HPTS liposome-based assay20, 35. Herein, we employed two recently reported complementary cationophore coupled assays19 to unambiguously confirm H+/Cl− symport facilitated by perenosins is a strictly inseparable process. Briefly, unilamellar vesicles were prepared with an internal solution of KCl (300 mM) and an external solution of K2SO4 (150 mM), buffered to pH 7.2 with potassium phosphate salts (5 mM). The naturally occurring cationophores valinomycin and monensin were either added alone, or in combination with a perenosin and the resulting chloride efflux, driven by a large chloride concentration gradient, was monitored using a chloride ISE. Valinomycin is selective for K+ and facilitates electrogenic transport which results in a net flow of charge across a lipid membrane, however, this does not directly affect the transmembrane pH gradient57. In contrast, monensin functions as an electroneutral transporter. Deprotonation of a carboxylic acid group occurs upon metal complexation allowing M+/H+ antiport (exchange) which results in an intra-vesicular pH change (illustrated in Fig. 6)58.

Compounds 1d and 2–6 were tested in the cationophore coupled assays and, in all cases, chloride efflux was observed when coupled to monensin and only a very small amount, <5%, of chloride efflux occurred in the presence of valinomycin. The K+/H+ antiport facilitated by monensin is coupled to the H+/Cl− cotransport facilitated by the receptors resulting in the overall KCl efflux observed. Therefore, the perenosins function as strict inseparable H+/Cl− cotransporters and display the same mechanism as prodigiosin, requiring protonation of the free imine-N2 to facilitate anion transport. The protonated perenosins cannot back-diffuse into the liposome resulting in the ‘strict’ HCl co-transport.

Selectivity mechanism studies

The selectivity for Cl−>H+/OH− transport of the perenosins was quantified using a modified NMDG-Cl assay18. Unilamellar vesicles loaded with HPTS (1 mM) and N-methyl-D-glucamine chloride (NMDG-Cl) (100 mM), buffered to pH 7 with HEPES (10 mM) were prepared. Free base NMDG (0.5 M) was added increasing the extravesicular solution to pH 8, and the dissipation of this gradient following the addition of receptors was monitored. Dissipation of pH gradient indicates H+/Cl− symport (or the functionally equivalent OH−/Cl− exchange) mediated by perenosins. Gramicidin D (Gra) is a proton channel added to facilitate H+ transport. The addition of oleic acid (OA), a naturally occurring fatty acid found in cell membranes, allowed the investigation of the enhancement in H+/Cl− symport, in the presence of oleic acid. Anion transporters can assist the fatty acid flip-flop mechanism by taking on the role of a flippase for fatty acid anions via binding to the anionic carboxylate head group, masking the charge and allowing rapid passage of the head group through the lipid bilayer. This in turn accelerates the natural proton transport mediated by fatty acids in biological membranes59.

For all cases EC50 values were obtained from dose-response experiments and subsequent Hill analysis (see Supplementary Information). Chloride selectivity was then quantified to give a selectivity factor (SG), shown in Table 2, using the ratio of the EC50 values. From the previous cationophore coupled assays, it was evident that the capability of perenosins to facilitate electrogenic chloride transport is negligible, therefore an absence in selectivity for Cl−>H+/OH− was expected. The selectivity for chloride (SG) is calculated by dividing EC50 in absence of Gra by EC50 (Gra), and the SG factors obtained for compounds 1d and 2–6 are all marginally greater than one which indicates a minor increase in the rate of transport in the presence of Gra, attributed to a small degree of Cl−>H+/OH− selectivity. Factor of enhancement (FOA) in the overall H+/Cl− cotransport rate facilitated by OA is calculated by dividing EC50 in absence of Gra or OA by EC50 (OA), a value greater than one suggests the receptor can assist the flip-flop of OA− increasing pH dissipation. FOA values for the overall H+/Cl− cotransport rate were calculated for compounds 1d and 2–6 in the presence of OA and are also shown in Table 2. Interestingly, compound 4 has a FOA value of 8.6, indicating a significant increase in the overall transport rate in the presence of fatty acids. This suggests that the receptor coupled fatty acid flip-flop is enhancing the overall rate of H+/Cl− cotransport (or OH−/Cl− exchange). This acceleration in transport rate also suggests that the rate limiting step in the NMDG-Cl assay for compound 4 is the back-diffusion of the neutral transporter across the lipid bilayer, due to the lower log P value. In the presence of fatty acids, an alternative pathway that involves the diffusion of the fatty acid complex and free fatty acids could replace this step. Conversely, compounds 2, 3 and 5 have FOA values lower than one, indicating slower H+/Cl− symport rate in the presence of fatty acids. This may be likely due to the competitive binding of the receptors to the carboxylate head group so inhibiting chloride binding and transport.

Other mechanistic studies

To further probe other aspects of transmembrane transport mechanism of perenosins, a variety of studies were employed. A Cl−/NO3 − antiport assay using an ISE method, as described previously in Cl−/NO3 − antiport at pH 7.2, was performed in POPC:cholesterol (7:3) vesicles; the addition of cholesterol decreases the fluidity and permeability of the membrane60 and this resulted in a lower percentage efflux facilitated by perenosins 1d and 2–8 compared to the same assay done in POPC vesicles. Sulfate is a highly hydrophilic, strongly hydrated anion that is difficult to transport over lipid bilayers61; using the Cl−/NO3 − antiport assay and changing the external solution to Na2SO4 from NaNO3 showed no Cl− efflux upon addition of 1d and 2–8, indicating that sulfate transport is not facilitated by the perenosins. Using the same conditions for sulfate transport, and adding a bicarbonate pulse at 120 s to give a final external bicarbonate concentration of 40 mM, allowed observation of potential Cl−/HCO3 − antiport elicited by 1d and 2–8; no Cl− efflux was detected, presumably due to the basicity of bicarbonate anion preventing protonation of the imine-N2. Finally, preincorporation of 1d and 2–6 into POPC vesicles followed by performing a similar NMDG-Cl assay, and subsequent dilution of the resulting lipid solution, investigated how perenosins behave in a lipid membrane environment. The H+/Cl− cotransport rates at different lipid concentrations remained unchanged, indicating that the perenosins remained in the lipid bilayer over the course of the transport experiments rather than partitioning into the aqueous phase. See Supplementary Information for more details of results and experimental details of these assays.

In Vitro Studies

Cell viability studies

Preliminary cell viability studies on the first generation perenosins conducted at the University of Southampton suggested cytotoxic selectivity for cancerous cells (MDA-MB-231 and MCF-7) over non-cancerous cells (MCF-10A)35. Herein, we carried out further in vitro studies at the University of Barcelona for the full library of perenosins. Single dose screening for cell viability using the methylthiazoletetrazolium (MTT) colorimetric assay over five different cell lines (cancer cells MCF-7, MDA-MB-231, A549 and SW620, and non-cancerous human cells MCF-10A) was conducted. After exposure to the perenosins (10 µM) for 24 h, interestingly, all perenosins exhibited selective cell viability decrease in human breast adenocarcinoma cell line (MDA-MB-231) over the other cell lines (see Supplementary Information).

Encouraged by these single dose screening results, dose response MTT assays were performed for the first time with compounds 2–8 and repeated for compounds 1a–e, on MDA-MB-231 cells to obtain IC50 values (inhibitory concentration of 50% of cell population) systematically at 24 h and 72 h incubation periods, see Table 3. Among compounds 1a–1e, 1d (being the best transporter in liposome studies) is the most potent after 24 h exposure time. Additionally, perenosin 1d remains the most potent compound after 72 h exposure time, and there is a slight decrease in cell viability by a factor of 1.2–1.7 for all the first generation of perenosins. Interestingly, compounds 2–8 exhibited an apparent structure-activity relationship for the onset of cytotoxic effect. There is negligible effect in cell viability (IC50 > 100 µM) after 24 h incubation time for perenosins with alkyl substituents at R2 (2–4, 6 and 8). However, there is a significant late onset of affecting the cell viability after 72 h, with 6 being the most potent perenosin (IC50 (72 h): 4.2 µM) in MDA-MB-231 cells. While perenosin 5, with pentyl substituent at R3, has similar cell viability effects as 1d, bis-indole perenosin 7 with pentyl substituent at R1 has a moderate degree of late onset by a factor of 7.

Table 3 IC50 values (µM, with error from three repeats in brackets) of first and second generations of perenosins 1a–e and 2–8 on the human breast adenocarcinoma (MDA-MCB-231) cell line with 24 h and 72 h exposure time.
Full size table

Cell-death mechanism studies

The specific late onset of affecting cell viability induced by perenosins is interesting, and to the best of our knowledge this phenomenon has not been observed for other anion transporters. Furthermore, this is governed by the unique structure-activity relationship displayed when bearing alkyl substituents at the R2 position (2–4, 6 and 8). However, the apparent potency (IC50) of these perenosins from the MTT assay does not correlate well with the anion transport efficacy from liposome-based studies. To investigate this further, Annexin-V was used as a marker with flow cytometry analysis to evaluate apoptosis in MDA-MB-231 cells after treatments with 2 (best transporter) and 6 (most potent) at their respective IC50 values. It is known that synthetic transporters can disrupt ion homeostasis in cells and induce cell death via an apoptotic pathway3, 22, 32, 63, 64.

As shown in Fig. 7 (see Supplementary Information for details), perenosin 2 clearly induces cell death by apoptosis, with a higher population of early apoptotic cells after 48 h exposure time and then transitioning to greater proportion of cells in late apoptosis after 72 h. Unexpectedly, the most potent perenosin 6 (from MTT cell viability assay) resulted in slight increase in apoptotic cells at the IC50 value after 48 h and 72 h. To confirm 6 induces apoptosis in cells, a higher concentration dose of 6 (10 µM) was used and demonstrated an increase in both early and late apoptotic cell populations. From this Annexin-V assay, we conclude that the best transporter 2 is more cytotoxic in inducing cell apoptosis, in contrast to the most potent compound 6 from the MTT assay. Therefore, the IC50 value of compound 6 obtained from MTT cell viability studies may be due to a combination of cell cycle arrest and apoptosis. This also suggests that other cellular effects can be induced by 6 such as DNA intercalation which disturbs the normal cell cycle, and not solely disruption of ion homeostasis. It should be mentioned that perenosins hydrolyse steadily (as shown in our preliminary investigation in aqueous buffer (see Supplementary Information)), due to the presence of the imine moiety.

Figure 7
figure 7

Results of Annexin-V flow cytometric assay on MDA-MB-231 cells, showing the distributions of cells present in the early and late apoptotic populations when treated with IC50 values of perenosins 2 (left) (11 µM) and 6 (right) (4 µM).

Full size image

Conclusions

The design of anion transporters often involves having a suitable alkyl chain to achieve optimal log P; however, we have also demonstrated the importance of positioning the alkyl chains to significantly enhance the anion binding and transport properties of the perenosins. While perenosin 1d emerged as the best compound from the original report (EC50 (ISE) 0.122 mol% and (HPTS) 0.0107 mol%), compound 2 shows a remarkable enhancement in the transmembrane anion transport (EC50 (ISE) 0.0043 mol% and (HPTS) 0.00062 mol%). This was achieved by changing the pentyl chain from position R1 to R2, to introduce the encapsulation/dehydration effects of the anion binding site by an alkyl chain. By using the cationophore-coupled ISE and gramicidin-coupled HPTS assays, the perenosins have been confirmed to function predominantly by the electroneutral mechanism of strictly inseparable H+/Cl− symport, in a similar fashion to prodigiosin. Additionally, the Cl−>H+/OH− selectivity as well as their transport capabilities in the presence of fatty acids have been investigated.

In vitro studies of our full library of perenosins demonstrated a decrease in cell viability of the human breast adenocarcinoma (MDA-MB-231) cell line over other non-cancerous and cancer cells. More interestingly, the new perenosins (2–4, 6 and 8) bearing an alkyl substituent at the R2 position serendipitously exhibited a distinctive late onset of cytotoxicity effect observed between 24 h and 72 h from the cell viability studies using MTT assays. Additionally, the cell death mechanism was evaluated using Annexin-V assays to confirm the apoptotic pathway as well as possible cell cycle arrest mechanism. The excellent anion transport efficacy in liposomes and cytotoxicity observed towards human breast adenocarcinoma cells demonstrated that the perenosins are potentially suitable candidates as anticancer therapeutics.