Biological interaction of living cells with COSAN-based synthetic vesicles

Cobaltabisdicarbollide (COSAN) [3,3′-Co(1,2-C2B9H11)2]−, is a complex boron-based anion that has the unusual property of self-assembly into membranes and vesicles. These membranes have similar dimensions to biological membranes found in cells, and previously COSAN has been shown to pass through synthetic lipid membranes and those of living cells without causing breakdown of membrane barrier properties. Here, we investigate the interaction of this inorganic membrane system with living cells. We show that COSAN has no immediate effect on cell viability, and cells fully recover when COSAN is removed following exposure for hours to days. COSAN elicits a range of cell biological effects, including altered cell morphology, inhibition of cell growth and, in some cases, apoptosis. These observations reveal a new biology at the interface between inorganic, synthetic COSAN membranes and naturally occurring biological membranes.

arrested proliferation and growth. Using cell proliferation in 4 day culture as an assay, we calculated ED 50 values of between 99-157 mM ( Figure 1B, Table 1).
For some cells and growth conditions, however we observed additional effects. When HEK293 cells were plated at low density, and hence had a higher proliferation rate, they began to bleb after 8 hours of treatment ( Figure 1C), having the appearance of cells entering apoptosis 14 with large numbers of cells dying at 24 hours. If HEK293 cells grown under these conditions were treated with both COSAN and the caspase-3 inhibitor Ivachtin to block apoptosis, cells remained viable for 24 hours ( Figure 1C), and fully recovered if COSAN was then washed out and cells grown for a further 24 hours ( Figure 1D). Finally, if HEK293 cells grown at low density were treated with COSAN for 5 hours, 3 hours before obvious apoptosis, and then COSAN was washed out, cells recovered and began to grow and proliferate normally ( Figure 1C and 1D).
In contrast, HeLa cells grown under the same conditions showed no evidence for blebbing or cell death. Instead, they took on an (Wash out at 5 hrs) following removal after 5 hours and further growth for 43 hours in fresh medium; (1caspase 3 inhibitor) or with COSAN plus caspase 3 inhibitor (100 nM Ivachtin) for 24 hours, followed by wash out and further growth for 24 hours in fresh medium, without COSAN or caspase 3 inhibitor. Data is plotted as percentage of untreated cell cultures, mean 6 SD for 3 independent experiments. (E) HeLa cells show an unusual, highly vacuolated morphology within the perinuclear cytoplasm when grown in COSAN for 24 hours. Following wash out of COSAN at either 5 or 12 hours the cells revert to the morphology seen in untreated cells (control).
www.nature.com/scientificreports SCIENTIFIC REPORTS | 5 : 7804 | DOI: 10.1038/srep07804 unusual, highly vacuolated appearance after 24 hours ( Figure 1E). When COSAN was applied to HeLa cells for 5 hours and then removed, cells reverted to the same morphology as untreated cells ( Figure 1D), indicating again that COSAN-mediated effects were reversible. Even when cells were exposed to COSAN for 12 hours and then washed free for a further 12 hours, the majority of cell showed no vacuolation, and those cells with vacuoles showed only minimal effects compared to 24 hour exposure. We conclude that of the cells examined apoptosis is restricted to HEK293 cells, and other cell types show different cellular changes not associated with cell death. As these additional effects require long-term exposure, greater than 5 hours, and are reversible if COSAN is removed before this time, we conclude that COSAN is not cytotoxic in the short term.
To probe the effect of COSAN on cell growth without these longerterm confounding effects, we switched the investigation to Dictyostelium discoideum cells, a eukaryotic cell that is evolutionarily close to animal cells, but lacks caspase-mediated apoptosis 15,16 . We found that these cells were much more sensitive to COSAN, with an ED 50 of 2.6 6 0.29 mM, and total arrest of proliferation occurring at 4 mM (Figure 2A-C). Under phase contrast microscopy, treated cells retained the refractile appearance of living cells, but rounded up and often exhibited a single large protrusion ( Figure 2A). Consistent with a lack of cytotoxicity, cells rapidly reverted to the untreated state following COSAN removal (Figure 2A and 2B). Due to the morphological change, cells became loosely attached to the underlying substratum. Cell detachment however cannot explain the growth arrest, as untreated Dictyostelium cells grow well in shaking suspension ( Figure 2C). As seen for mammalian cells, cell rounding was also observed when COSAN was added in phosphate buffer (KK2) or water. In growth medium Dictyostelium cells remained viable for long periods in COSAN. To date, we have found no limits for Dictyostelium, with cultures remaining viable for more than 1 month (37 days) in 10 mM COSAN, and recovering after removal. In addition, cells were viable and grew well in COSAN-free culture, even after prolonged exposure to 500 mM of COSAN, almost 200 times the ED 50 value ( Figure 2D).
Modification of COSAN alters its cellular potency. Using the more sensitive Dictyostelium cells, we tested a range of chemical variations and modifications of COSAN ( Figure 3). Substitution of cobalt for iron as the central metal (FESAN) made no major difference to compound potency. However, methylation of COSAN, or linking two COSAN clusters via a PEG chain decreased cell potency ( Figure 3). In contrast, addition of iodine to make di-iodo-COSAN (I 2 -COSAN) increased potency (Table 1; Figure 3 and 4A). Dictyostelium cell recovery after I 2 -COSAN treatment took longer than COSAN and was dose-dependent, suggesting a stronger interaction with its cellular targets ( Figure 4B). When tested on mammalian cell cultures, again I 2 -COSAN was more potent than COSAN ( Figure 4C). In addition, we tested the influence of different biocompatible cations: salts of COSAN, Li 1 and Na 1 (ED 50 values were 2.5 and 2.6 mM, respectively); salts of FESAN, Li 1 and Na 1 (ED 50 values were 3.2 and 3.1 mM, respectively); salts of I 2 -COSAN, H 1 and Na 1 (ED 50 values of both salts was 1.8 mM), but found no differences in effects on cells. On this basis, all biological tests were carried out with Na 1 as the cation of the COSAN anion and its derivatives.
We had found no, or very little, effect of COSAN on the growth of bacteria. In contrast, I 2 -COSAN arrested growth of the two bacterial species we tested, E. coli and K. pneumoniae, ( Figure 4D). Interestingly, we found that E. coli was significantly more sensitive to I 2 -COSAN than K. pneumoniae. Growth arrest was reversed by I 2 -COSAN removal or dilution to below its ED 50 value ( Figure 4E).

Direct visualization of I 2 -COSAN within cells.
To investigate why I 2 -COSAN was more potent than the original COSAN molecule, we used micro-Raman spectroscopy to compare uptake and release of both molecules. As previously reported, the distinctive vibrational peak at 2570 cm 21 of B-H bonds provides a means of direct visualization and quantification of COSAN and I 2 -COSAN in living HEK293 cells 13 , showing that both COSAN and I 2 -COSAN accumulates inside cells ( Figure 5). In fact, we found that both COSAN and I 2 -COSAN reach intracellular concentrations substantially greater than in the external medium (PBS). This is illustrated well in the case of I 2 -COSAN, where an extracellular 2 mM concentration of I 2 -COSAN, a concentration too low for normal detection, accumulated in 30 minutes to a concentration of approximately 25 mM within cells, a greater than 10-fold rise ( Figure 5A, supplementary Figure. S1 online). Although both molecules possessed the ability to accumulate against the concentration gradient, we observed that I 2 -COSAN reached an intra-cellular concentration 3-4 times that of COSAN ( Figure 5B). When COSAN or I 2 -COSAN was then removed from the medium, they initially remained within the cell and then were gradually lost over the preceding hours. However 4 hours after COSAN or I 2 -COSAN removal from the external environment, there was more retained inside the cell than originally applied to the outside, and I 2 -COSAN was still detectable within cells after 4 days ( Figure 5B). These observations indicate that the rate of entry of both COSAN and I 2 -COSAN is greater than the exit rate, and can explain the intracellular accumulation.
Although absolute concentration of I 2 -COSAN inside the cells was higher than COSAN, in both cases the concentration 4 hours after removal was 20% of the total peak concentration ( Figure 5B). This indicated that COSAN and I 2 -COSAN exited cells at the same rate, and did not explain why intracellular I 2 -COSAN was higher than COSAN. To investigate this further, we measured the octanol-water Partition Coefficient (P) as an indication of lipophilicity (supplementary Figure. S2 online), and found that I 2 -COSAN has a partition value (P 5 151.0) approximately 3.5-fold higher than COSAN (P 5 43.7). This study indicates that I 2 -COSAN possesses significantly higher lipophilicity than COSAN.
The ability to accumulate and measure I 2 -COSAN within cells that are incubated in I 2 -COSAN concentrations below the detection threshold offers a strong signal contrast with which to image the subcellular distribution. As previously reported for treatment with higher COSAN concentrations, cells treated with 2 mM I 2 -COSAN showed that although present within both the cytoplasm and nucleus, I 2 -COSAN was not homogeneously distributed, with some cytoplasmic regions showing higher concentrations than others ( Figure 5A). Currently, we do not have sufficient spatial resolution to correlate these higher concentration regions with known sub-cellular structures. It is very clear, however, that there was no substantial accumulation in the plasma membrane, consistent with our previous biophysical observations that COSAN passed directly through artificial membranes 12 .

Discussion
The interaction of COSAN and related molecules with synthetic and biological membranes was described in previous reports 12,13 , and here we have explored the biological interactions with living cells. COSAN and I 2 -COSAN form vesicles in aqueous solution 4 and have the unusual property of transiting lipid bilayers with zero order kinetics 12 , where rate is independent of the initial concentration. Previously, we observed fusion and transit of COSAN vesicles through the lipid membrane of the liposomes 12 , and COSAN entry into living cells 13 . In the experiments reported here, we show that COSAN entry into cells neither disrupts the plasma membrane nor is immediately cytotoxic. This is consistent with the previous biophysical experiments 12 that demonstrated that the boron-rich anion COSAN and its I 2 -COSAN derivative cross through cell-free artificial lipid bilayer synthetic membranes, either prokaryotic (DPhPC: 1,2-diphytanoyl-sn-glycero-3-phosphocholine) or eukaryotic (1,2dioleoyl-sn-glycero-3-phosphocholine, DOPC). We found that for most eukaryotic and prokaryotic cell cultures, cells remain viable with prolonged COSAN treatment, but have arrested cell growth and proliferation, indicating a cytostatic, not  cytotoxic, effect. When COSAN is washed from the cells, this cytostatic effect is reversed and they resume growth. However, under some conditions, additional cell effects were observed: proliferating HEK293 cells exhibiting apoptosis and HeLa cells forming large intracellular vacuoles. These additional, cell-specific events require exposure of many hours, contrasting to the morphological changes, followed by growth arrest, that occurs rapidly after COSAN exposure in all eukaryotic cells. These observations suggest that the interaction of COSAN with cells is more than a simple physicochemical process and may entail more active interactions with intracellular cell signalling pathways that control proliferation, death or membrane dynamics. The nature of these interactions is currently unknown. With our current imaging techniques we cannot determine whether separate COSAN vesicles form within cells as seen in artificial liposome experiments or whether COSAN forms more complex structures within the cytoplasm, such as hybrid COSAN and phospholipid membranes. We note however the association between membrane fluidity, cell proliferation arrest and apoptosis seen with other small molecules, such as methyl jasmonate 17 , and it is possible that all the biological effects we observe could arise due to membrane interactions. Alternatively, COSAN may mediate its effects due to protein interactions either by enzyme inhibition via active site binding, such the modified COSAN molecules that inhibit enzymes 18,19 , or more general interference by binding to protein surfaces. More detailed analysis is required to resolve these mechanistic possibilities.
The effects of COSAN are seen in a broad range of conditions, spanning different counter ions (H 1 , Li 1 and Na 1 ) and cell media and buffers. In addition, the core effects on cell morphology and cell growth arrest are conserved over many cell types, including Dictyostelium and bacteria, although the sensitivity to COSAN and its derivatives varies widely over different cell types. Changing the structure of the COSAN had a large effect on the potency of its effect, with diCOSAN structures having no or very little effect, whereas addition of iodine (I 2 -COSAN) enhances its biological effects.
To investigate why I 2 -COSAN has higher potency than COSAN, we used micro-Raman spectroscopy to monitor both COSAN and I 2 -COSAN within living cells. We found that both molecules exit cells at a slower, but equal, rate in comparison to entry. This leads to accumulation within the cell against the concentration gradient, however I 2 -COSAN accumulates to a higher level. Currently, we do not know whether COSAN accumulation reaches a maximum concentration, however the cell biological effects described in this paper occur at much lower concentrations of COSAN or I 2 -COSAN than used for imaging, and so there is plenty of capacity for both to accumulate to a level with significant effects.
A key difference between COSAN and I 2 -COSAN is the approximate 3.5 times higher lipophilicity of the I 2 -COSAN molecule, which suggests a higher affinity for the lipid environment. This difference is also consistent with observations on artificial membranes 12 , where the rate-limiting factor for COSAN and I 2 -COSAN translocation is partitioning between lipid and water phase, and the slower permeation rate seen for I 2 -COSAN could result from its greater affinity for lipids. Importantly, we do not see either COSAN or I 2 -COSAN accumulation within the plasma membrane of cells further demonstrating that COSAN anions do not accumulate within the lipid membrane phase.
A valuable feature of the inorganic boron-based COSAN membranes reported here is that they are not present in nature and therefore unlikely to be further modified or degraded by cellular enzymes. Consistent with this, we have observed no chemical modification of COSAN or I 2 -COSAN in the cell systems reported here (see supplementary Figure S3 online). This biologically inert feature may offer new opportunities for drug design and molecular delivery systems. Recent work demonstrates that functionalized I-COSAN vesicles show good biodistribution and can be imaged throughout whole mice 20 , and we would predict that they would not show major toxicity if concentrations are maintained below 100 mM. COSAN, and other borane clusters, can be coupled to bioactive molecules, such as inhibitors, siRNA and protein ligands [21][22][23][24] , and may have the potential to carry these molecules across the plasma membrane and into cells. COSAN vesicles could also be used to encapsulate water soluble or hydrophilic compounds, however it is currently unclear how the topology of the COSAN vesicles changes as they merge with lipid bilayer membranes and whether this would deliver into the cell or disperse on the cell surface.
As COSAN can accumulate to high concentrations in living cells, this may enhance potential therapeutic effects. For example, its cell accumulation may be beneficial for Boron neutron capture therapy (BNCT) proposed for cancer therapy [25][26][27] , or other clinical usages. We also note the differential sensitivity between cell state and cell species. In tissue culture, we observed cell death in proliferating cells following prolonged COSAN treatment ( Figure 1C), however confluent monolayers appeared unaffected. If this is translated to the in vivo context, then it may be possible to target rapidly growing cells. Dictyostelium amoebae show a much higher sensitivity to COSAN and its derivatives. If this higher sensitivity to COSAN were also present in other unicellular eukaryotes, it may offer an amoebicidal agent to target protozoan and amoebozoan pathogens.
Placed in the context of earlier papers, these observations demonstrate that COSAN is able to accumulate within cells without affecting membrane integrity. Once inside the cell, COSAN is not immediately cytotoxic, but is cytostatic over the long term, and cells recover following its removal. Specific biological effects, however, vary with cell type and conditions. These results reveal novel biology interactions at the interface of biological and synthetic membranes.  31 were synthesized as described in the literature. The water-insoluble compounds were dissolved in a mixture of acetonitrile/water (50550) and passed repeatedly through a cation exchanging resin, which had been loaded previously with the desired cation