• Alkylphosphocholines;
  • Antitumor agents;
  • Exploratory data analysis;
  • Hemolytic potential;
  • Molecular modeling


  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Methods
  5. 3 Results and Discussion
  6. 4 Conclusions
  7. Acknowledgements
  8. Supporting Information

Alkylphosphocholines (APC) are promising antitumor agents, which have the cellular membrane as primary target; however, red blood cell damage limits their wide therapeutic use. A variety of APC analogs has been synthesized and tested showing less hemolytic effect than the class prototype, Miltefosine (HePC). In this work, chemometric methods were applied to a set of 34 APC derivatives to identify the most relevant structural and molecular features of hemolytic activity. The APC derivatives were divided into three groups: (i) N-methylpiperidine and N-methylmorpholine derivatives with a long alkyl chain or flexible cyclopentadecyl rings, displaying a hemolytic rate of 17 %; (ii) adamantyl and cyclopentadecyl derivatives, showing an average hemolysis of 39 %; and, N,N,N-trimethylammonium, trans-N,N,N-trimethylcyclohexanamine, and trans-N,N,N-trimethylcyclopentanamine derivatives, whose average hemolysis was 41 %. The findings suggested that the presence of either bulky cationic head groups, or rings such as adamantyl and cyclohexyl, primarily increases the hemolysis of compounds with eleven atoms in the alkyl chain. Moreover, the macrocyclic cyclopentadecyl seems to be important to the hemolytic potential especially of compounds with five carbon atoms in the alkyl chain. Regarding linear carbon chain derivatives with no ring substitution, less bulky cationic head groups seem to favor hemolysis. Thus, in order to design more potent and less toxic APC antitumors, the reported structural/molecular patterns should not be included in their structure.

1 Introduction

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Methods
  5. 3 Results and Discussion
  6. 4 Conclusions
  7. Acknowledgements
  8. Supporting Information

Alkylphosphocholines (APC) represent a promising class of antitumor agents derived from alkyl-lysophospholipids (ALP), which are, in turn, related to endogenous lysophospholipids.1 Edelfosine (ET-18-OCH3, 1-O-octadecyl-2-O-methyl-rac-glycero-3-phosphocholine) is an ALP derivative and its antineoplastic activity against various tumor cells encouraged the search for new synthetic phospholipids analogs.1 Structure-activity relationship studies of a variety of ALP analogs showed that antiproliferative effects do not strongly depend on the position and chemical character of the substituent in the glycerol backbone. Miltefosine (hexadecylphosphocholine, HePC), which contains the minimal structural requirements for antineoplastic activity, such as a long alkyl chain and a phosphocholine moiety, was synthesized as the APC class prototype.1,2 Different from other DNA-targeting antineoplastic agents, there are evidences that the APC class interferes with membrane rafts, phospholipid metabolism, cholesterol vesicular transport and homeostasis, survival biochemical pathways (such as the Akt-mTOR) and proteins involved in signal transduction, such as protein kinase C, phospholipase C and phospholipase D. However, a clear mechanism of action has not been established yet.3,4

Miltefosine (HePC) is currently employed for topical treatment of cutaneous metastases of breast carcinoma. It can also be combined with other chemotherapeutic agents, resulting in less myelotoxicity.5 However, the use of APC as anticancer agents has been limited owing to their severe side effects, especially gastrointestinal toxicity when orally administrated, and hemolysis and thrombophlebitis when intravenously injected.2,3 APC are amphiphilic compounds and in aqueous solutions, at concentrations higher than the critical micellar concentration (CMC), they form micelles which fuse to form cylindrical aggregates on a long-ranged hexagonal lattice, known as hexagonal I (HI) phase, or “normal topology”. In this topology, the polar heads are pointing to the aqueous solution and the alkylic chains are in the micellar core.[6] The inverse topology, or hexagonal II phase (HII), is formed in principle when the amphiphile concentration is increased beyond the point where lamellar phases are formed, or when the micelles are in a hydrophobic environment, such as the bilayer core. Furthermore, the hydrocarbon chains are pointing to the hydrophobic environment, while the polar heads are in the micellar core.6 At hemolytic concentrations, the formation of non-lamellar HII can give rise to regions that destabilize the bilayer creating an eventual site for the initiation of cell disruption.[7] The interaction of APC with biological membranes can induce other biophysical changes, such as membrane shape changes, lipid flip-flop, vesiculation and solubilization.7 At low concentrations, however, surfactants such as APC can ward off hypotonic hemolysis of cells, although the mechanism underlying this anti-hemolytic effect is not fully understood. It is supposed that by intercalating into the lipid bilayer, surfactants can expand the cellular membrane, allowing the cell to swell to a larger volume before it lysis, being this a protective mechanism. The protection against hypotonic hemolysis seems to be the result of a nonspecific interaction between lipids of the bilayer and surfactant molecules.8

In order to improve safety and tolerability, APC analogs were previously designed and synthesized by other groups.[9,10] The influence of cycloalkane rings (especially the presence of 4-alkylidenecyclohexyl and cycloalkylidene groups in alkoxyethyl and alkoxyphosphodiester ether lipids) on the APC antiproliferative activity against eight human and animal cell lines [PC3, human prostate adenocarcinoma; MCF7, human breast adenocarcinoma; A431, human epidermoid carcinoma; HeLa, human cervical adenocarcinoma; PC12, pheochromocytoma (rat adrenal); U937, human promonocytic leukemia; K562, human chronic myelogenous leukaemia; CHO, ovary biopsy (Chinese hamster)] has been investigated.9 Moreover, the hemolytic activity of the ring-substituted ether phospholipids synthesized in that study was also evaluated.10 The majority of the compounds tested exhibited significant cytotoxic activity, which depended on the size and position of the ring with respect to the phosphate moiety, as well as the head group characteristic. This effect, however, varied among the cell types tested.9 The most potent compound exhibited broad-spectrum anticancer activity, comparable to HePC’s activity against MCF7 and U937 cell lines; and, superior to HePC’s activity against PC3, A431, HeLa and CHO cells, apart from being non-hemolytic. Thus, the findings suggested that the exploited molecular modifications played an important role in the hemolytic activity.9 Herein, structural differences of part of this set of compounds were investigated by means of molecular modeling and chemometric methods, so as to discriminate which molecular properties (related to chemical structure/molecular modifications) would be important to improve anticancer activity and/or reduce hemolytic activity.

Exploratory data analysis is a chemometric approach which comprises two unsupervised pattern recognition methods: hierarchical cluster analysis (HCA)11,12 and principal component analysis (PCA).11,12 HCA is a multivariate method for calculating and comparing distances between pairs of samples or variables, grouping data into clusters with common attributes. Its primary purpose is to emphasize clusters and patterns of the investigated data, displaying them on a two-dimensional plot, known as dendrogram. PCA is a data compression method based on correlations among variables or descriptors. Its purpose is to reduce the dimensionality of a data set, grouping correlated variables and replacing original descriptors with a new set, called principal components (PC), onto which all data is projected. Principal components are variables completely uncorrelated, obtained by simple linear combinations from the original variables. Moreover, they contain most of the variability of the original data in a lower dimensional space.12

In this work, exploratory analysis of a set of APC analogs9,10 was performed to identify the most relevant descriptors or molecular properties (thermodynamic, electronic, steric, topological, and structural) responsible for compounds’ discrimination, and further establish correlations with the biological data, in this case, the hemolytic activity. Besides giving better insight into APC analogs’ cytotoxicity profile and hemolytic mechanism, the findings can also be helpful in designing more potent and less toxic antitumor agents.

2 Methods

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Methods
  5. 3 Results and Discussion
  6. 4 Conclusions
  7. Acknowledgements
  8. Supporting Information

2.1 Biological Data Set

The data set comprising thirty-four APC analogs (including HePC) was selected from the works of Papazafiri et al.9 and Calogeropoulou et al.10 The chemical structures and biological data of the investigated compounds are listed in Table 1. In both studies, the hemolytic potential was measured in percentage of hemolysis (%Hem) at 100 µM by the same experimental protocol, using EDTA-preserved peripheral blood from healthy volunteers.9,10

Table 1. Structures of the alkylphosphocholine analogs with their respective percent hemolysis (%Hem) at 100 µM, selected from P. Papazafiri et al.9 and T. Calogeropoulou et al.10.
Thumbnail image of
Thumbnail image of
Thumbnail image of
Thumbnail image of
Thumbnail image of
Thumbnail image of
Thumbnail image of
Thumbnail image of
Thumbnail image of
Thumbnail image of
Thumbnail image of
Thumbnail image of
Thumbnail image of
Thumbnail image of
Thumbnail image of
Thumbnail image of
Thumbnail image of
Thumbnail image of
Thumbnail image of
Thumbnail image of
Thumbnail image of
Thumbnail image of
Thumbnail image of
Thumbnail image of
Thumbnail image of
Thumbnail image of
Thumbnail image of
Thumbnail image of
Thumbnail image of
Thumbnail image of
Thumbnail image of
Thumbnail image of
33 (HePC)
Thumbnail image of
Thumbnail image of

2.2 Building up the Three-Dimensional (3D) Molecular Models

The 3D molecular models of all compounds were built up using the HyperChem 7.51 software.13 The cartesian coordinates of Edelfosine were retrieved from the Cambridge Structural Database (CSD; entry code DONZAH)14 and employed as the starting geometry for constructing all molecules. The energy minimization of each ligand was carried out in MM+ force field,15 without any constraint, with the Polak-Ribière conjugate gradient algorithm (HyperChem 7.51). Partial atomic charges were calculated using PM316 semiempirical method, implemented in HyperChem 7.51.13 MOLSIM 3.2 package17 was also used for geometry optimization procedure applying steepest descent and, subsequently, conjugated gradient method. The convergence criterion adopted was 0.01 kcal/mol. The energy-minimized structures were used as inputs for molecular dynamics (MD) simulations, also performed in MOLSIM 3.2 software.17

2.3 Molecular Dynamics (MD) Simulations and Descriptors Calculation/Selection

MD simulations were performed to generate a conformational ensemble profile (CEP) of each ligand. All MD simulations were carried out during 1 ns [1 000 000 steps; step size 0.001 ps (1 fs)] and at 310 K for each ligand. The output trajectory files were recorded every 20 steps, generating 50 000 conformers for each molecule. After the equilibration, the lowest-energy conformer was selected from the CEP, and the hydration shell model proposed by Hopfinger[18] was used to estimate the solvation energy contribution. The hydrogen bonding intramolecular energy contribution was also computed for the lowest-energy conformation of each ligand. At this point, thermodynamic properties were obtained for each molecule. The selected conformers were energy-minimized before calculating other molecular properties (independent variables or descriptors) using the appropriate software. Descriptors related to steric, electronic, hydrophobic, topological, geometric, and structural properties were calculated employing the following packages: HyperChem 7.51, Marvin,19 Gaussian 03W.20 Detailed information regarding all descriptors, methods, and respective packages used are listed in Table 2.

Table 2. Calculated descriptors for the investigated set of compounds
ChelpGNElectronicN2 electrostatic potential chargesHF/6-31G (d,p) ChelpG22,23Gaussian 03W for Windows20
ChelpGP P1 electrostatic potential charges  
EHOMO Energy of the highest occupied molecular orbital (HOMO)  
ELUMO Energy of the lowest unoccupied molecular orbital (LUMO)  
µx X component of dipole moment  
µy Y component of dipole moment  
µz Z component of dipole moment  
µtot Total dipole moment  
EstrThermodynamicStretching energy contributionMD simulations (see Method section)MOLSIM 3.217
Ebend Bending energy contribution  
Etors Torsional energy contribution  
E1,4 Lenard-Jones or 1–4 type interactions energy contribution  
EvdW van der Waals energy contribution  
Eel Electrostatic energy contribution  
EHb Hydrogen bonding contribution  
Esolv Solvation energy contribution  
Etot Total potential energy (sum of all energy contribution)  
SAaproxSteric intrinsic/stericApproximate Å2 surface area24HyperChem 7.5 for Windows13
SAgrid Surface area obtained by GRID method25 
Vsol Solvent accessible volume  
V_vdW van der Waals volume  
VNPsol Non-polar portion solvent accessible volume  
VNPvdw Non-polar portion van der Waals volume  
VPsol Polar portion solvent accessible volume  
VPvdw Polar portion van der Waals volume  
θN2StructuralTorsional angle N2-C12-C11-O4  
θO4 Torsional angle O4-P1-O3-C8  
θP1 Torsional angle P1-O3-C8-C9  
Dist_P1TopologicalInteratomic distance between P1 and N2 atoms  
CnumSteric/geometricNumber of carbons in the hydrophobic portion  
ClogPHydrophobicCalculated partition coefficient (n-octanol/water)26 
MRSteric/hydrophobicMolar refractivity  
MassIntrinsicMolecular mass  
ClogPionHydrophobicCalculated partition coefficient (n-octanol/water) of ionic species26Marvin
ClogPnion Calculated partition coefficient (n-octanol/water) of nonionic species  
ClogD7.4Hydrophobic/electronicApparent partition coefficient in different pH systems: 7.4, 6.5, 5.0, and 1.5  
α3DElectronic/steric3D geometry (Thole) to calculate the polarization tensor values27 
PSA Polar surface area28 
MMFF94ThermodynamicConformational energy using Merck Molecular Force Field29 
DreGeometricDreiding energy (conformers stability)[30] 
SAsolSteric/geometricSolvent accessible surface area24 
AminGeometricConformer’s minimum projection area based on van der Waals radius25 
Amax Conformer’s maximum projection area based on van der Waals radius  
Rmin Conformer’s minimum radius area  
Rmax Conformer’s maximum radius area  
PlattTopologicalPlatt index31 
Randic Randic index32 
Balaban Balaban index33 
Harary Harary index[34] 
Szeged Szeged index[35] 
Wiener Wiener index36 
HyWiener Hyper Wiener index  
PWiener Wiener polarity index  

After the descriptors calculation, a matrix containing 34 rows (number of samples or compounds) and 62 columns (calculated molecular descriptors or independent variables) (block X) was generated (see Supporting Information; the data were split into five tables, A to E). The %Hem values were expressed as logarithmic scale and represent the dependent variable or y vector, which is the last column in Table E (Supporting Information). All samples (rows) were previously randomized before running the exploratory data analyses. A filter using the Pearson correlation coefficient value of 0.2 was formerly used as cut off for selecting the calculated independent variables. Visual inspection of scatter plots of the biological data (%Hem) versus each calculated descriptor was also performed, and variables presenting uniform dispersion/distribution, as well as linear correlation tendencies in relation to the biological data, were primarily selected for composing the final matrix (input for the exploratory analyses, HCA and PCA), since the Pirouette 3.11 program21 only uses linear functions. Variables displaying non-uniform distribution (dispersion) were discarded in this step. The final block X exhibited the same number of rows (34 compounds or samples), but different number of columns (10 descriptors or variables selected plus %Hem).

2.4 Exploratory Data analysis (HCA and PCA)11,12

HCA and PCA methods were carried out in Pirouette 3.11 program.21 The autoscaling procedure was applied as a preprocessing method in order to normalize the different orders of magnitude of the calculated variables.12

The complete linkage method and Euclidean distance were considered in HCA. The distances between samples or variables were calculated according to Equation 1.

  • equation image(1)

The multivariate distance dkl between two sample vectors, k and l, is determined by computing differences in each of the m variables. M is the order of the distance, and here represents the Euclidean distance (M=2). The distance values were transformed into a similarity matrix whose elements are the similarity indices (similaritykl=1−dkl/dmax, where dmax is the largest distance in the data set). The similarity scale ranges from zero (dissimilar samples or variables) to one (identical samples or variables), and the larger the similarity index, the smaller the distance between any pair of samples or variable.11,12 The findings are expressed as a dendrogram (see Figure 3 in Section 3 Results and Discussion).

PCA is a multivariate method whose principal components contain most of the variability of the data set in a much lower dimensional space. PC1 or factor 1 is defined along the direction of maximum variance of the whole data set, whereas PC2 or factor 2 is on the direction that describes the maximum variance of the subspace orthogonal to PC1. The subsequent components are taken orthogonally to those previously chosen, and describe the maximum of the remaining variance. Once redundant information is removed, only the first few PCs are necessary to describe most of the information contained in the original data set. The data matrix X (I×J), corresponding to I molecules and J descriptors, is broken down into two matrices, T and L, in such a way that X=TLT. The T matrix, known as the score matrix, represents the position (classification) of the compounds on the new coordinate system where the PCs are the new axes. Scores are integral to exploratory analysis because they show intersample relationships. L is the loadings’ matrix whose columns describe how the new axes (PCs) are built from the old axes. It also gives insight into the variables importance, that is to say, which variables contribute more or less to each PC or factor.11,12 In this exploratory data analysis, PCA was run up to nine factors. The Pirouette 3.11 software displays many entities that can help to explore the relationships between samples, find samples outliers, choose the optimal number of factors and make decisions about variables to be excluded. The outliers’ diagnosis, for example, was performed through the Mahalanobis distance (Equation 2),21,37 which is a distance computed from its k factor (PC) score:(2)

  • equation image(2)

In Equation 2, S is the score covariance matrix and equation image is the mean score vector. Assuming that the Mahalanobis distance is normally distributed, a critical value (MDcrit) can be determined from the chi squared distribution with k degrees of freedom. If the sample’s Mahalanobis distance exceeds MDcrit, that sample might be an outlier.

3 Results and Discussion

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Methods
  5. 3 Results and Discussion
  6. 4 Conclusions
  7. Acknowledgements
  8. Supporting Information

PCA findings are shown in Figure 1. According to the factor selection, the first two PCs were responsible for explaining 85.01 % of the total variance from the original data (Figure 1A). The score plot revealed that the two first PCs divided the investigated set into three main groups, which are displayed as A (compounds 3, 4, 10, 11, 12 and 19), B (compounds 5, 7, 9, 15, 16, 17, 18, 20, 21, 23, 25, 26, 28, 29 and 30), and C (compounds 1, 2, 6, 8, 13, 14, 22, 24, 27, 31, 33 and 34) (Figure 1B). Compound 32 was more distant from the rest of data set and was not included in any group.

thumbnail image

Figure 1. PCA findings: (A) factor selection (plot and table), (B) score plot, and (C) loading values (plot and table).

Download figure to PowerPoint

The loadings plot and table (see Figure 1C) provides information on descriptors distribution across the PCs, as well as which calculated properties influenced more the compounds’ separation. In PC1, intrinsic/steric (Vsol, V_vdW), steric/hydrophobic (MR), electronic (α) and topological (Randic and Harary indices) properties or descriptors have higher loading values. On the other hand, in PC2, topological (Dist_P1, Platt index) and hydrophobic (ClogP) descriptors influenced more the compounds separation. Steric and topological parameters are related to the molecular shape and play an important role in interaction processes.

Regarding amphiphilic compounds, it is noteworthy to mention that hemolysis occur through mechanisms involving surfactant-membrane interaction and membrane disruption/solubilization. The presence of micelles is important in this mechanism, which involves micellar dissolution and partition into membrane, phase transition between lamellae and micelles and decrease in the size of micelles.38 Micellar aggregation depends on the space occupied by the hydrophilic and hydrophobic groups of the surfactants. The volume and length occupied by the hydrophobic portion in the micellar core, and the cross-sectional area occupied by the hydrophilic portion at the micelle-solution interface are important parameters that influence aggregation, size and shape of micelles.6,39 It is generally accepted that the amount of surfactant required to solubilize a membrane increases with the ease of forming micelles, expressed by the CMC values.40

The outliers diagnosis plot, which considers sample residual versus Malahanobis distance, shows that except to compounds 32 and 33 (both highly hemolytic), the samples were in the residual’s threshold (solid line; see Figure 2). The threshold is based on a 95 % confidence interval set internally in Pirouette 3.11.21 However, just exploratory findings from PCA are not sufficient to consider compounds 32 and 33 as outliers. Then, according to the calculated molecular properties, which are directly dependent on the chemical structure, compounds 32 and 33 can be considered slighlty different from the rest of data set. The ether phospholipid, compound 32, has a more constrained tropane-derived head group. This structural feature has been reported as responsible for its lower cytotoxic effect.10 Miltefosine (33) has no cycle/ring in any moiety of its chemical structure, increasing the degrees of freedom (higher flexibility).

thumbnail image

Figure 2. Plot of sample residual versus Mahalanobis distance. The sample residual threshold (light gray line) is based upon a 95 % probability limit (set internally in Pirouette software).21

Download figure to PowerPoint

Complementary findings were obtained for both methods, HCA and PCA, reinforcing the separation pattern, as can be seen in the samples/compounds dendrogram (Figure 3A). Also, the topological properties, as Platt index and Dist_P1, were grouped in the same cluster of the biological data (see Figure 3B, dendrogram of variables). These two properties presented also high loading values in PCA method.

thumbnail image

Figure 3. HCA findings: (A) Dendrogram of samples; (B) Dendrogram of variables.

Download figure to PowerPoint

Regarding the dendrogram of samples, using a similarity cursor of 0.5 (dashed line in Figure 3A), the same three main groups were formed, corresponding to cluster A (68 % of similarity), B (65 % of similarity), and C (55 % of similarity). Once more, compound 32 was not grouped anywhere, considering that similarity cursor. This compound, however, shares 44 % of similarity with the molecules grouped in cluster B. Cluster A was basically constituted by N-methylpiperidine and N-methylmorpholine derivatives with long alkyl chain or flexible cyclopentadecyl rings. Regarding the hemolytic rate of those compounds, the average is 17 %±11 %. Cluster B comprises a wide variety of structures, many of which with bulkier substituents as adamantyl or cyclopentadecyl in the lipophilic portion. This cluster grouped all the 11-adamantyl derivatives. It is noteworthy that compounds 7, 18, 20, 23, and 25 form a sub-cluster sharing 81 % of similarity, and they have similar polar head group moieties (N-methylpiperidine or N-methylmorpholine). Additionally, those molecules have fairly moderate hemolytic rates. In this case, the cationic moiety seemed to influence the position of this sub-cluster, since the compounds have substantially different alkyl chains (compounds 18 and 25 have cyclopentadecyl ring, whereas compounds 7 and 23 have an adamantyl portion). These compounds share 74 % of similarity with 16, 26, and 30. Compound 29 presented similarity index of 0.69 with those compounds. Apart from 26 and 30, the rest of the molecules are N-methylpiperidino derivatives. This sub-cluster comprises short alkyl chain compounds with a cyclopentadecyl ring (16, 18, 25, 26, 29, 30).

Compounds 5, 9, 15, 17, 21, and 28 have a similarity index of 0.70. This sub-cluster did not include any N-methylpiperidine or N-methylmorpholine derivatives, but rather N,N,N-trimethylammonio, trans-N,N,N-trimethylcyclohexanamine, and trans-N,N,N-trimethylcyclopentanamine. Corroborating the results seen in cluster A (compounds 7, 18, 20, 23, and 25), the cationic head seemed to play a role in grouping those molecules. This set of compounds shares 65 % of similarity with the previous sub-cluster. Regarding hemolysis, cluster B essentially comprises molecules whose hemolytic rate is not so high, and the average for this group is 39 %±22 %. Cluster C was formed at 55 % of similarity, and it essentially comprises N,N,N-trimethylammonium, trans-N,N,N-trimethylcyclohexanamine and trans-N,N,N-trimethylcyclopentanamine derivatives, and just one N-methylpiperidine entity (compound 8). Compounds 6, 8, 13, 14, 22, 27, 31, and 34, in cluster C, presented a high similarity index (79 %). Molecules 1, 2, 24 and 33 (Miltefosine) also presented a similarity index well above 50 %, which is considered high (78 %), and all have N,N,N-trimethylammonium as cationic head group. Miltefosine (33) and compound 24 shared 89 % of similarity. Including the carbon atoms of the cycle, compound 24 has a tail with 17 carbons (eleven carbon atoms plus six-membered ring, whereas Miltefosine’s tail has 16 carbons). The same similarity index was shared by molecules 1 and 2 (89 %). These two molecules differ from one another only by the presence of a double bound adjacent to the ring (compound 2). Regarding hemolysis, cluster C grouped rather highly hemolytic compounds (average hemolytic rate of 41 %±36 %), especially compounds 27, 31, 33 and 34. The high standard deviation of this group when compared to clusters A and B denotes the presence of compounds with high hemolytic rates. The difference between this group of compounds and other highly hemolytic molecules, found in cluster B, is the presence of bulkier substituents in the alkyl chain of the latter (26, 28, 29 and 30), instead of linear chain (33 and 34) or smaller six-membered rings (27 and 31).

As far as hemolysis is concerned, it can be considered that highly hemolytic compounds (hemolysis >60 %) are essentially distributed across clusters B and C, and different structural features can be observed (Table 3). In cluster B, highly hemolytic compounds have adamantyl or cyclopentadecyl rings as substituents in an alkyl chain with eleven carbon atoms, except for 26, 29 and 30. When the cationic moiety is taken into account, compounds in B present trans-N,N,N-trimethylcyclohexanamine (26 and 28), trans-N,N,N-trimethylcyclopentanamine (30), and N-methylpiperidine (29) groups. Highly hemolytic compounds in cluster C have also, at least, eleven carbon atoms in the alkyl chain (27, 31, 33 and 34), without any cycle (33 and 34), or with a small six-membered ring (27 and 33) at the end of the alkyl chain. The molecules mentioned before present the following cationic head group moieties: trans-N,N,N-trimethylcyclohexanamine (31), trans-N,N,N-trimethylcyclopentanamine (27 and 34), and N,N,N-trimethylammonium (33). Moreover, the presence of an unsaturation in the carbon next to the ring (26, 27, 28, 29, 30 and 31) seems to be important to the hemolytic potential.

Table 3. Summary of the main structural features and their effect on hemolysis.
 Ring substitutions at the end of the alkylic chainUnsaturationHead group
 Bulky rings (cyclodecyl or cyclopentadecyl)Less bulky rings (adamantyl, cyclohexyl)Unsaturation in the carbon next to the ringBulky cationic head group (trans-N,N,N-trimethylcyclohexanamine,­ trans-N,N,N-trimethylcyclopentanamine or­ N,N-dimethyltropane)N,N,N-trimethyl ammoniumN-methylpiperidineN-methylmorpholine
Compounds with eleven carbons in the alkyl chain + ring[DOWNWARDS ARROW]Hemolysis (10, 11, 14, 19)[UPWARDS ARROW]Hemolysis (27, 28, 31, 32) Exception: 17[UPWARDS ARROW]Hemolysis (27, 28, 31, 32)[UPWARDS ARROW]Hemolysis (27, 28, 31, 32) Exception: 17[DOWNWARDS ARROW]Hemolysis (9, 11, 14, 21, 24)[DOWNWARDS ARROW]Hemolysis (7, 8, 19)[DOWNWARDS ARROW]Hemolysis (10, 23)
Linear saturated chain with no rings[DOWNWARDS ARROW]Hemolysis (13trans-N,N,N-trimethylcyclohexanamine) [UPWARDS ARROW]Hemolysis (34trans-N,N,N-trimethylcyclopentanamine)[UPWARDS ARROW]Hemolysis (33)Not observedNot observed
Compounds with 5 carbons in the alkyl chain+ring[UPWARDS ARROW]Hemolysis (when combined with a bulky cationic head group) (26, 30) [DOWNWARDS ARROW]Hemolysis (when combined with a less bulky cationic head group) (1, 2, 6, 16, 18, 22, 25) Not observed[UPWARDS ARROW]Hemolysis (26, 29, 30)[UPWARDS ARROW]Hemolysis (when combined with bulky ring substitutions) (26, 30) [DOWNWARDS ARROW]Hemolysis (when combined with bulky ring substitutions) (1, 2, 6, 22)[UPWARDS ARROW]Hemolysis (29 – unsaturated version of 16)[DOWNWARDS ARROW]Hemolysis (18, 25)
Compounds with cycloalkoxy moiety in the middle of the alkylic chainCompounds with low hemolytic rates, regardless of the substitution (3, 4, 5, 12, 15, 20)

In addition to what was mentioned in the previous section, when eleven carbon alkyl chains with bulkier rings or six-membered rings are associated with an N,N,N-trimethylammonium cationic head, the hemolytic effect is reduced (9, 11, 14, 21, 24). In this regard, it seems important that either linear saturated chains be combined with this cationic moiety (33), or that those molecules with eleven carbon atoms and rings at the end of the alkyl chain present bulkier cationic head groups (trans-N,N,N-trimethylcyclohexanamine, trans-N,N,N-trimethylcyclopentanamine, or N-methylpiperidine). The trans-N,N,N-trimethylcyclohexanamine and trans-N,N,N-trimethylcyclopentanamine cationic head groups seem to be necessary to elevate the hemolytic potential in molecules with eleven carbons in the alkyl chain, since compounds 7, 8, 9, 10, 19, 23 and 24 are just moderately hemolytic and have N-methylpiperidine, N-methylmorpholine or N,N,N-trimethylammonium as cationic head groups. The exception is compound 17 and, in this case, the N,N,N-trimethylcyclopentanamine seems to have a negative effect on hemolysis (as opposed to compound 28). It is also noteworthy that N-methylmorpholine derivatives are not highly hemolytic (hemolysis rate < 60 %), indicating that this substitution reduces red blood cell damage.

Another pattern is observed for molecules 26, 29 and 30. They were grouped in cluster B with other short-tailed compounds having cationic head groups bearing a ring and a cyclopentadecyl ring in the tail (18 and 25), instead of the N,N,N-trimethylammonium moiety and cyclodecyl ring in the tail (1, 2, 6 and 22 – cluster C). For those compounds with shorter alkyl chains (five carbon atoms), it seems important that a bulkier cationic head group, rather than an N,N,N-trimethylammonium portion (compound 22 is only fairly hemolytic), be present. The cyclopentadecyl ring, in combination with bulkier cationic head groups, also seems to produce more hemolytic compounds, since molecules 1 and 2 are just slightly hemolytic. As discussed before, the N-methylmorpholine portion has an apparently negative effect on hemolysis, and compound 25 (which has similar structure to 29), presents hemolytic rate of approximately 45 %.

The recognition pattern methods applied herein also allowed identifying structural features related to the hemolytic potential of the investigated set, since molecular properties are directly dependent on the chemical structure (Table 3). From the findings here obtained, highly hemolytic compounds should present bulkier cationic head groups. For those compounds containing eleven carbon atoms in the alkyl chain, macrocyclic rings (cyclodecyl or cyclopentadecyl) reduce hemolysis. On the other hand, the less bulky adamantyl and cyclohexyl rings can be responsible for improving the hemolytic potential (such as compounds 27, 28 and 31). Compound 32, which was not grouped anywhere, combines an alkyl chain containing eleven carbon atoms, an adamantyl moiety, and a bulky N,N-dimethyltropane cationic head group, resulting in a highly hemolytic molecule (82 %). Moreover, for compounds with five carbon atoms in the alkyl chain, a bulky cyclopentadecyl ring seems to be important for the hemolytic rate (26, 29 and 30). Another important aspect is the presence of an unsaturation adjacent to the ring at the end of the alkylic chain. As far as compounds without any ramification or ring substitution in the alkyl chain are concerned, less bulky cationic head groups seem to be favorable to hemolysis (33 and 34). However, compound 13 (analog of 34), which has a trans-N,N,N-trimethylcyclohexanamine moiety (larger than trans-N,N,N-trimethylcyclopentanamine), is just a fairly hemolytic molecule.

4 Conclusions

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Methods
  5. 3 Results and Discussion
  6. 4 Conclusions
  7. Acknowledgements
  8. Supporting Information

Considering the above-mentioned findings, this study provided useful information on which structural features of APC class play an important role in hemolysis and, in a broader sense, surfactant interaction with biological membranes. The fact that bulkier cationic head groups lead to more hemolytic compounds is quite important to propel the design of novel less toxic compounds. In this sense, a less hemolytic compound would have small cationic head groups, such as N,N,N-trimethylammonium, and steric bulk rings at the end of the alkylic chain (Figure 4). Also, another option would be the design of short-tailed compounds containing small cationic head groups and large macrocyclic rings at the end of the alkylic chain (Table 3). In both cases, the combination of a small cationic head group and bulkier substituents in the alkylic chain seems to be necessary to reduce hemolysis. The design of antitumor surfactants with less steric bulk cationic portions, such as N,N,N-trimethylammonium, is under way.

thumbnail image

Figure 4. Proposition of a chemical scaffold for less hemolytic alkylphosphocholine derivatives.

Download figure to PowerPoint

Further studies are required to better understand alkylphoscholines’ interaction with biological membranes, but the current work sheds light on their toxicity and how structure is related to hemolysis. By avoiding this harmful side effect, this promising class could be vastly applied to treat cancer.


  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Methods
  5. 3 Results and Discussion
  6. 4 Conclusions
  7. Acknowledgements
  8. Supporting Information

This research was supported by grants from the Coordination for Higher Level Graduate Improvements (CAPES/Brazil), the National Council for Scientific and Technological Development (CNPq/Brazil) and the State of São Paulo Research Foundation (FAPESP/Brazil). The authors thank the Chem21 Group, Inc., for the academic license of MOLSIM 3.2 software.

  • 1
    H. Brachwitz, C. Vollgraf, Pharmacol. Ther. 1995, 66(1), 3982.
  • 2
    V. Jendrossek, K. Hammersen, B. Erdlenbruch, W. Kugler, R. Krügener, H. Eibl, M. Lakomek, Cancer Chemother. Pharmacol. 2002, 50(1), 7179.
  • 3
    W. J. Van Bitterswijk, M. Verheij, Biochim. Biophys. Acta 2013, 1831(3), 663674.
  • 4
    J. M. Jiménez-López, P. Ríos-Marco, C. Marco, J. L. Segovia, M. P. Carrasco, Lipids Health. Dis. 2010, 9(33), 110.
  • 5
    M. C. Georgieva, S. M. Konstantinov, M. Topashka-Ancheva, M. R. Berger, Cancer Lett. 2002, 182(2), 163174.
  • 6
    R. G. Laughlin, The Aqueous Behaviour of Surfactants, Academic Press, London, 1996.
  • 7
    S. Scheirer, S. V. P. Malheiros, E. Paula, Biochim. Biophys. Acta. 2000, 1508(1–2), 210234.
  • 8
    B. Isomaa, H. Hägerstrand, G. Paatero, A. C. Engblom, Biochim. Biophys. Acta 1986, 860(3), 510524.
  • 9
    P. Papazafiri, N. Avlonitis, P. Angelou, T. Calogeropoulou, M. Koufaki, E. Scoulica, I. Fragiadaki, Cancer Chemother. Pharmacol. 2005, 56(3), 261270.
  • 10
    T. Calogeropoulou, P. Angelou, A. Detsi, I. Fragiadaki, E. Scoulica, J. Med. Chem. 2008, 51(4), 897908.
  • 11
    K. R. Beebe, R. J. Pell, M. B. Seasholtz, Chemometrics: A Practical Guide, Wiley, New York, 1998.
  • 12
    M. M. C. Ferreira, J. Braz. Chem. Soc. 2002, 13(6), 742753.
  • 13
    HyperChem(TM) Professional 7.51, Hypercube, Inc., 1115 NW 4th Street, Gainesville, Florida 32601, USA.
  • 14
    I. Pascher, S. Sundell, H. Eibl, K. Harlos, Chem. Phys. Lipids 1986, 39(1–2), 5364.
  • 15
    N. L. Allinger, J. Am. Chem. Soc. 1977, 99(25), 81278137.
  • 16
    J. J. P. Stewart, J. Comput. Chem. 1989, 10(2), 221264.
  • 17
    D. Doherty, MOLSIM: Molecular Mechanics and Dynamics Simulation Software. User’s Guide, Version 3.2, The Chem21 Group Inc, Lake Forest, IL, 1997.
  • 18
    A. J. Hopfinger, in Conformational Energies and Potential Functions (Ed: A. J. Hopfinger), Academic Press, New York, 1973, p. 71.
  • 19
    Marvin, ChemAxon, 2013;
  • 20
    M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb, J. R. Cheeseman, J. A. Montgomery, Jr., T. Vreven, K. N. Kudin, J. C. Burant, J. M. Millam, S. S. Iyengar, J. Tomasi, V. Barone, B. Mennucci, M. Cossi, G. Scalmani, N. Rega, G. A. Petersson, H. Nakatsuji, M. Hada, M. Ehara, K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, O. Kitao, H. Nakai, M. Klene, X. Li, J. E. Knox, H. P. Hratchian, J. B. Cross, V. Bakken, C. Adamo, J. Jaramillo, R. Gomperts, R. E. Stratmann, O. Yazyev, A. J. Austin, R. Cammi, C. Pomelli, J. W. Ochterski, P. Y. Ayala, K. Morokuma, G. A. Voth, P. Salvador, J. J. Dannenberg, V. G. Zakrzewski, S. Dapprich, A. D. Daniels, M. C. Strain, O. Farkas, D. K. Malick, A. D. Rabuck, K. Raghavachari, J. B. Foresman, J. V. Ortiz, Q. Cui, A. G. Baboul, S. Clifford, J. Cioslowski, B. B. Stefanov, G. Liu, A. Liashenko, P. Piskorz, I. Komaromi, R. L. Martin, D. J. Fox, T. Keith, M. A. Al-Laham, C. Y. Peng, A. Nanayakkara, M. Challacombe, P. M. W. Gill, B. Johnson, W. Chen, M. W. Wong, C. Gonzalez, J. A. Pople, Gaussian 03, Revision B.02, Gaussian, Inc., Wallingford CT, 2004.
  • 21
    Pirouette 3.11, 19902003, Infometrix, Inc.,Woodinville WA.
  • 22
    V. A. Fock, Z. Phys. 1930, 61, 126148.
  • 23
    G. A. Petersson, M. A. Al-Laham, J. Chem. Phys. 1990, 94(9), 60816091.
  • 24
    W. Hasel, T. F. Hendrickson, W. C. Still, Tet. Comput. Meth. 1988, 1(2),103116.
  • 25
    A. Gavezzotti, J. Am. Chem. Soc. 1983, 105(16), 52205225.
  • 26
    V. N. Viswanadhan, A. K. Ghose, G. R. Revankar, R. K. Robins, J. Chem. Inf. Comput. Sci. 1989, 29(3),163172.
  • 27
    K. J. Miller, J. Savchik, J. Am. Chem. Soc. 1979, 101(24), 72067213.
  • 28
    P. Ertl, B. Rohde, P. Selzer, J. Med. Chem. 2000, 43(20), 37143717.
  • 29
    T. A. Halgren, J. Comput. Chem. 1996, 17(5–6), 490519.
  • 30
    S. L. Mayo, B. D. Olafson, W. A. Goddard, J. Phys. Chem. 1990, 94(26), 88978909.
  • 31
    J. R. Platt, J. Phys. Chem. 1952, 56(3), 328336.
  • 32
    M. Randic, J. Am. Chem. Soc. 1975, 97, 66096615.
  • 33
    A. T. Balaban, Chem. Phys. Lett. 1982, 89, 399404.
  • 34
    D. Plavšić, S. Nikolić, N. Trinajstić, Z. Mihalić, J. Math. Chem. 1993, 12, 235250.
  • 35
    I. Gutman, Graph. Theory. Notes 1994, 27, 915.
  • 36
    H. Wiener, J. Am. Chem. Soc. 1947, 69(1), 1720.
  • 37
    P. C. Mahalanobis, J. Asiat. Soc. Bengal. 1930, 26, 541588.
  • 38
    H. Heerklotz, Q. Rev. Biophys. 2008, 41(3–4), 205264.
  • 39
    J. N. Israelachivili, in Aggregation of Amphiphilic Molecules into Micelles, Bilayers, Vesicles, and Biological Membranes (Ed: J. N. Israelachivili), Academic Press, London, 2011, pp. 246264.
  • 40
    P. S. Preté, K. Gomes, S. V. Malheiros, N. C. Meirelles, E. De Paula, Biophys. Chem. 2002, 97(1), 4554.

Supporting Information

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Methods
  5. 3 Results and Discussion
  6. 4 Conclusions
  7. Acknowledgements
  8. Supporting Information

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.


Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.