17β-carboxamide steroids - In vitro prediction of human skin permeability and retention using PAMPA...

Accepted Manuscript

17β −carboxamide steroids - in vitro prediction of human skin permeability andretention using PAMPA technique

Vladimir Dobričić, Bojan Marković, Katarina Nikolic, Vladimir Savić, SoteVladimirov, Olivera Čudina

PII: S0928-0987(13)00424-7DOI: http://dx.doi.org/10.1016/j.ejps.2013.10.017Reference: PHASCI 2908

To appear in: European Journal of Pharmaceutical Sciences

Received Date: 19 July 2013Revised Date: 30 September 2013Accepted Date: 23 October 2013

Please cite this article as: Dobričić, V., Marković, B., Nikolic, K., Savić, V., Vladimirov, S., Čudina, O., 17β−carboxamide steroids - in vitro prediction of human skin permeability and retention using PAMPA technique,European Journal of Pharmaceutical Sciences (2013), doi: http://dx.doi.org/10.1016/j.ejps.2013.10.017

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customerswe are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, andreview of the resulting proof before it is published in its final form. Please note that during the production processerrors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

http://dx.doi.org/10.1016/j.ejps.2013.10.017

http://dx.doi.org/http://dx.doi.org/10.1016/j.ejps.2013.10.017

��

�

17β−carboxamide steroids - in vitro prediction of human skin

permeability and retention using PAMPA technique

Vladimir Dobričića,*, Bojan Markovića, Katarina Nikolica, Vladimir Savićb, Sote Vladimirova,

Olivera Čudinaa

aDepartment of Pharmaceutical Chemistry, Faculty of Pharmacy, University of Belgrade,

Vojvode Stepe 450, 11000 Belgrade, Serbia

bDepartment of Organic Chemistry, Faculty of Pharmacy, University of Belgrade, Vojvode Stepe

450, 11000 Belgrade, Serbia

*Corresponding author. Address: Department of Pharmaceutical Chemistry, Faculty of

Pharmacy, University of Belgrade, Vojvode Stepe 450, 11000 Belgrade, Serbia. Tel.: +381 11

3951 335; fax: +381 11 3972 840.

E-mail address: [email protected] (V. Dobričić).

Abstract

In this paper, twenty-two 17β-carboxamide steroids were synthesized from five corticosteroids

(hydrocortisone, prednisolone, methylprednisolone, dexamethasone and betamethasone) in two

steps. The first step was periodic acid oxydation of these corticosteroids to corresponding

cortienic acids and the second step was amidation of thus obtained cortienic acids with esterified

��

�

L-amino acids. These compounds are potential soft corticosteroids with local anti-inflammatory

activity in the skin. PAMPA (Parallel Artificial Membrane Permeability Assay) was applied in

order to predict permeability and retention of these compounds in human skin. Comparison of

permeability and retention parameters between 17β-carboxamide steroids and corresponding

corticosteroids was performed. Compounds with significantly higher retention were identified

and the derivative that doesn't have significantly higher permeability was underlined. Molecular

structures of all compounds were optimized by use of Gaussian semiempirical/PM3 method.

Geometrical, thermodynamic, physicochemical and electronical molecular parameters of the

optimized structures were calculated and quantitative structure-property relationship (QSPR)

analysis was performed in order to explain permeability and retention of these compounds. ANN-

, PLS- and MLR-QSPR models were created. Quality of these models was evaluated by

commonly used statistical parameters and the most reliable models were selected. Analyzing

descriptors in the selected models, main molecular properties that influence permeability and

retention in the PAMPA artificial membrane were identified. Based on these data, further

structural modifications could be applied in order to increase retention without significant

increase of permeability, which can positively affect potential local anti-inflammatory activity of

these compounds. Selected QSPR models could be used as in silico tool for predicting human

skin permeability and retention of novel 17β−carboxamide steroids without performing PAMPA

experiments.

Key words: Soft corticosteroids; PAMPA; human skin permeability and retention; quantitative

structure-property relationship.

��

�

1. Introduction

Soft drugs are applied or administered at or near the site of action usually to produce local effect.

Metabolical deactivation of these drugs prevents undesired pharmacological activity or toxicity.

Presence of a metabolically sensitive moiety in the drug molecule enables its design and

prediction of the major metabolic pathway. As a result, the formation of undesired toxic, active,

or high-energy intermediates can be prevented. The soft nature of a drug is usually related to fast

hydrolytic degradation. However, if hydrolysis is too fast, poor activity may be obtained (Bodor

and Buchwald, 2006). Therefore, a soft drug must be stable enough to reach the receptor sites at

the target organ and to produce desired effect. By modifying the structure of a soft drug, it is

possible to change its pharmacokinetic profile, pharmacological activity and toxicity.

An example of a soft corticosteroid is loteprednol etabonate, an ester of cortienic acid (Bodor et

al., 1992a, 1992b; Druzgala et al., 1991a, 1991b). Cortienic acid is the major corticosteroid

metabolite and it lacks glucocorticoid activity. Apart from esters, amides of cortienic acids have

been synthesized and tested for glucocorticoid or antiglucocorticoid activity (Formstecher et al.,

1991; Manz et al., 1983, 1984; Rousseau et al., 1979). Several amides showed glucocorticoid

activity (inhibition of phytohaemagglutinin-mediated proliferation of lymphocites) (Manz et al.,

1983, 1984).

The parallel artificial membrane permeability assay (PAMPA) is a simple and rapid test intended

to estimate passive transport permeability (Kansy et al., 1998). In PAMPA, a 96-well filter plate

impregnated with a liquid artificial membrane is used to separate two compartments: donor

compartment (contains a buffer solution of compounds to be tested) and acceptor compartment

https://www.researchgate.net/publication/51335036_Physicochemical_High_Throughput_Screening_Parallel_Artificial_Membrane_Permeation_Assay_in_the_Description_of_Passive_Absorption_Processes?el=1_x_8&enrichId=rgreq-90d0e066-b372-4b1f-afc3-7e1f72087969&enrichSource=Y292ZXJQYWdlOzI1ODYzNTEyNTtBUzoxMDI1NDAzNTYwOTYwMDFAMTQwMTQ1ODkzNDQ2Nw==

��

�

(contains an initial fresh buffer solution) (Kansy et al., 1998; Ottaviani et al., 2006). Ottaviani et

al. (2006) tested different combinations of silicone oil and isopropyl myristate (IPM) and

concluded that the artificial membrane composed of 30% IPM and 70% silicone oil allows best

prediction of human skin permeability. Apart from the combination of silicone oil and IPM,

certramides (synthetic ceramide analogs) could also be used for the simulation of barrier

properties of human skin (Sinkó et al., 2009, 2012).

The aim of this study was to synthesize, structurally characterize twenty-two L-amino acid

amides of cortienic acids obtained from five corticosteroids (hydrocortisone, prednisolone,

methylprednisolone, dexamethasone and betamethasone) and to predict their permeability

through human skin and human skin retention using PAMPA technique. It is expected that these

compounds are metabolized in vivo to corresponding cortienic acid and L-amino acid. Therefore,

these derivatives are potential soft corticosteroids with local anti-inflammatory activity in the

skin. Geometrical, thermodynamic, physicochemical and electronical descriptors of these

compounds will be calculated and quantitative structure-property relationship (QSPR) study will

be performed in order to identify the most important physicochemical properties that influence

permeability and retention in the artificial PAMPA membrane. These data could be used in

further structural modifications of synthesized compounds in order to increase their retention

without significant increase of permeability, which can positively affect their potential local anti-

inflammatory activity.

�

�

https://www.researchgate.net/publication/51335036_Physicochemical_High_Throughput_Screening_Parallel_Artificial_Membrane_Permeation_Assay_in_the_Description_of_Passive_Absorption_Processes?el=1_x_8&enrichId=rgreq-90d0e066-b372-4b1f-afc3-7e1f72087969&enrichSource=Y292ZXJQYWdlOzI1ODYzNTEyNTtBUzoxMDI1NDAzNTYwOTYwMDFAMTQwMTQ1ODkzNDQ2Nw==

https://www.researchgate.net/publication/6994116_Parallel_Artificial_Membrane_Permeability_Assay_A_New_Membrane_for_the_Fast_Prediction_of_Passive_Human_Skin_Permeability?el=1_x_8&enrichId=rgreq-90d0e066-b372-4b1f-afc3-7e1f72087969&enrichSource=Y292ZXJQYWdlOzI1ODYzNTEyNTtBUzoxMDI1NDAzNTYwOTYwMDFAMTQwMTQ1ODkzNDQ2Nw==

https://www.researchgate.net/publication/40031050_A_PAMPA_Study_of_the_Permeability-Enhancing_Effect_of_New_Ceramide_Analogues?el=1_x_8&enrichId=rgreq-90d0e066-b372-4b1f-afc3-7e1f72087969&enrichSource=Y292ZXJQYWdlOzI1ODYzNTEyNTtBUzoxMDI1NDAzNTYwOTYwMDFAMTQwMTQ1ODkzNDQ2Nw==

��

�

2. Materials and methods

2.1. Chemicals

Hydrocortisone was purchased from Alfa Aesar (Karlsruhe, Germany). Prednisolone,

methylprednisolone, dexamethasone, betamethasone and methyl ester of L-alanine were

purchased from Tokio Chemical Industry (Tokyo, Japan). Methyl ester of L-glycine, N-

hydroxybenzotriazole (HOBt) and N,N-dimethylformamide (DMF) were purchased from Sigma

Aldrich (Steinheim, Germany), whereas 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDC),

triethylamine (TEA), ethyl ester of L-glycine, ethyl ester of β-alanine and ethyl ester of L-

phenylalanine were purchased from Acros Organics (Geel, Belgium). Tetrahydrofuran (THF),

isopropyl myristate, IPM (>95%) and silicone oil (DC 200) were purchased from Fluka Chemie

GmbH (Bruch, Switzerland), while NaH2PO4·H2O and Na2HPO4 were from Merck (Darmstadt,

Germany). Chloroform and methanol were purchased from JT Baker (Loughborough, UK). Silica

gel for column chromatography was purchased from Merck (Darmstadt, Germany) and silica gel

for preparative thin-layer chromatography was purchased from Sigma-Aldrich (Steinheim,

Germany). Acetonitril Chromasolv HPLC purity (Sigma-Aldrich Chemie GmbH, Steinheim,

Germany) and deionised water (TKA water purification system, Niederelbert, Germany) were

used throughout this study.

�

2.2. General procedure for the synthesis of cortienic acids

��

�

Cortienic acids were synthesized in high yields (90-97.5%) by periodic acid oxydation of

corticosteroids according to the literature procedure (Bladh et al., 2010). The reaction scheme is

presented in Fig. 1.

< Fig. 1>

2.3. General procedure for the synthesis of 17β−carboxamide steroids

Amides were synthesized according to the literature procedure (Anthes et al., 2009). Cortienic

acids (0.14 mmol, 1 eq) were dissolved in DMF (2 ml) at room temperature. Subsequently,

corresponding amino acid (0.17 mmol, 1.19 eq), EDC (0.21 mmol, 1.5 eq), HOBt (0.21 mmol,

1.5 eq) and TEA (38.5 μl, 1.98 eq) were added. Reaction mixtures were stirred at room

temperature overnight. The reaction mixtures were evaporated to dryness under reduced pressure.

The residues were dissolved in the mixture of chloroform and methanol and purified by column

chromatography. Mobile phases used for column chromatography purification were

chloroform/methanol 99:1 (v/v) (A), chloroform/methanol 98.5:1.5 (v/v) (B) and

chloroform/methanol 98:2 (v/v) (C). The purity of collected fractions was determined by TLC

and HPLC. Unsatisfactorily pure fractions were evaporated to dryness, dissolved in methanol and

repurified using preparative thin-layer chromatography. Mobile phases used for preparative thin-

layer chromatography were chloroform/methanol/glacial acetic acid 95:5:1 (v/v/v) (D) and

chloroform/methanol 95:5 (v/v) (E). Finally, purified products were recrystallized in the mixture

of water and methanol to obtain white or light-gray crystalline solids.

�

�

The synthesized compounds were structurally characterized by determing melting points and by

spectroscopic methods (UV, IR, NMR, MS-TOF and MS/MS). Melting points were determined

on Boetius PHMK 05 apparatus (Germany). UV spectra were recorded using Evolution 300 UV-

VIS spectrometer (Thermo Fisher Scientific, UK). IR spectra were recorded using FT-IR

spectrometer Nicolet iS10 (Thermo Fisher Scientific, Madison, WI, USA). NMR spectra were

recorded on NMR BRUKER AVANCE III DMX 500 (Bruker Biospin GmbH, Rheinstetten,

Germany). 13C NMR data are presented in Supplementary table 1. Exact masses were determined

using liquid chromatograph Agilent Technologies 1210 combined with mass detector Agilent

6210 Time-of-Flight (Agilent Technologies, Palo Alto, CA, USA) and LTQ Orbitrap XL FT

Mass Spectrometer (Thermo Fisher Scientific, Bremen, Germany). MS/MS analyses were

performed using TSQ Quantum Access MAX triple quadrupole mass spectrometer (Thermo

Fisher Scientific, San Jose, USA), equipped with heated electrospray ionization source (HESI).

2.4. PAMPA test

The membrane solution (30% isopropyl myristate–70% silicone) was prepared in n-hexane (35%

v/v). Phosphate buffer (pH 5.5, ionic strength: 20 mM) was prepared by dissolving 2.6460 g

NaH2PO4·H2O and 0.1177 g Na2HPO4 in water (1000 ml). Test solutions were prepared by

dissolving about 1 mg of corticosteroids and their 17β-carboxamide derivatives in DMSO (1.25

ml) in 25 ml volumetric flasks and filling up to the mark with the phosphate buffer. Acceptor

solution was prepared as 5% DMSO solution in the buffer.

The PAMPA test was performed according to the literature procedure (Markovic et al., 2012;

Ottaviani et al., 2006) in hydrophobic PVDF 96-well filter plates (Multiscreen IP Filter Plate 0.45

�

�

μm) from Millipore (Bedford, MA, USA). Compounds were tested in triplicates at iso-pH

conditions. Each well of the donor plate was coated with the membrane solution (17 μl). Next, in

each well of the acceptor plate, acceptor solution (300 μl) was transferred and covered by the

donor plate to create a PAMPA sandwich. Finally, in each well of the donor plate test solutions

(300 μl) were transferred. The donor plate was covered to prevent evaporation from donor wells

and the whole system was set up to interact with the vibratory mixer. After 7 h, the PAMPA

sandwich was disassembled. Concentrations of all tested compounds in each well, as well as in

starting solutions, were determined using LC-MS method.

�

2.4.1. LC- MS method

TSQ Quantum Access MAX triple quadrupole mass spectrometer, equipped with heated

electrospray ionization source (HESI), was utilized for mass spectrometric detection. The MS

detector was set to detect and quantify corresponding ions (m/z) under positive HESI mode for

hydrocortisone, methylprednisolone, dexamethasone and betamethasone or under negative HESI

mode for other tested compounds. The chromatographic analysis was performed using Accela

Thermo Scientific system consisted of Accela Pump and Autosampler. The mobile phase

composition was acetonitrile : 0.1% formic acid (60 : 40, v/v for determination of phenylalanine

derivatives and 50:50, v/v for determination of other compounds). All runs were performed using

Zorbax Eclipse XDB-C18 column (100 x 4.6 mm, 3.5 μm particle size), which was maintained at

25 oC. The temperature of autosampler was 6 oC. The injection volume was 10 μl, while the flow

rate of the mobile phase was set at 500 μl/min. For data acquisition, Xcalibur 1.2 software

(Thermo Fisher Scientific, San Jose, USA) was used.

��

�

2.4.2. Calculations of PAMPA parameters

The values of permeability coefficients–logPe, permeation parameters–CA(t)/CD(0) and retention

factors–R of tested compounds were calculated using equations (1) and (2) (Avdeef, A., 2003;

Ottaviani et al., 2006). R can be defined as the mole fraction retained in the membrane and in the

microplates (i.e., filters and plate materials):

��

)0(

)(

)0(

)(1

D

A

D

A

D

D

C

tC

V

V

C

tCR −−= ��

��

⎥⎥⎦

⎤

⎢⎢⎣

⎡⎥⎦

⎤⎢⎣

⎡⎟⎟⎠

⎞⎜⎜⎝

⎛−

+−⎟⎟⎠

⎞⎜⎜⎝

⎛+−

−=)0(

)(

)1(1log

)(

303.2loglog

D

A

D

DA

DA

A

LAG

D

C

tC

RV

VV

VV

V

ttA

VPe ��

�

VA- volume in the acceptor wells (ml)

VD- volume in the donor wells (ml)

A-filtration area (cm2)

t-incubation time (s)

tLAG -steady–state time (s)

CD(t)-concentration of the compound in donor well at time t (μM)

CA(t)-concentration of the compound in acceptor well at time t (μM)

CD(0)-concentration of the compound in donor well at time 0 (μM)

Steady–state time (tLAG) which is needed to saturate the membranes in PAMPA is relatively short

compared to the total permeation time (about 20 min with unstirred plates) (Avdeef et al., 2001).

�

https://www.researchgate.net/publication/11669446_Drug_absorption_in_vitro_model_filter-immobilized_artificial_membranes_2_Studies_of_the_permeability_properties_of_lactones_in_Piper_methysticum_Forst?el=1_x_8&enrichId=rgreq-90d0e066-b372-4b1f-afc3-7e1f72087969&enrichSource=Y292ZXJQYWdlOzI1ODYzNTEyNTtBUzoxMDI1NDAzNTYwOTYwMDFAMTQwMTQ1ODkzNDQ2Nw==

��

�

2.5. Calculation of geometrical, thermodynamic, physicochemical and electronical molecular

parameters and statistical analysis

Geometry of all compounds was minimized by use of Gaussian semiempirical/PM3 method,

included in Chem3D Ultra 9.0.1 program (CambridgeSoft Corporation, 2005). Geometrical,

thermodynamic, physicochemical and electronical molecular parameters of the optimized

molecular structures were calculated in Chem3D Ultra 9.0.1, MarvinSketch (ChemAxon, 2008)

and Dragon software (Talete srl, 2010). Using Dragon, Chem3D Ultra 9.0.1 and MarvinSketch,

more than 4500 molecular parameters were initially calculated. Molecular parameters calculated

in MarvinSketch and Chem 3D Ultra 9.0.1 are presented in Table 1.

< Table 1>

Quantum chemically based reactivity molecular parameters, such as chemical potential (μ),

electronegativity (χ), hardness (η), global softness (S) and electrophilicity index (ω) were

calculated from HOMO and LUMO energies (Filipic et al., 2013; Iczkowski and Margrave, 1961;

Parr et al., 1989).

Subsequently, intercorrelation between calculated molecular parameters was tested. Pairs of

molecular parameters with intercorrelation higher than 0.99 (for PLS analysis) or higher than

0.90 (for ANN and MLR analysis) were examined and those with stronger influence on

dependent variables (logPe and R) were retained for modeling. After the completion of

intercorrelation test, 279 molecular parameters were used for further MLR(logPe), ANN(logPe),

MLR(R) and ANN(R) analysis, whereas 654 molecular parameters were used for further

PLS(logPe) and PLS(R) analysis.

��

�

STATISTICA package with the neural network module was used for stepwise MLR and ANN

modeling (StatSoft Inc., 1998). The Soft Independent Modeling of Class Analogy SIMCA P+

12.0 program (Umetrics AB, 2008)� was used for the PLS analysis (Eriksson et al., 2001;

Umetrics AB, 2008). In the PLS analysis, variables were selected according to the variable

importance in the projection (VIP) parameter (Umetrics AB, 2008).

Single factor ANOVA was performed in Microsoft Office Excel (Microsoft Corporation, 2010)

in order to identify compounds with significantly higher retention than starting corticosteroids.

3. Results and discussion

3.1. Selection and synthesis of 17β-carboxamide steroids

The affinity for glucocorticoid receptor depends on the nature of 17β side chain. According to X-

ray crystal structure studies (Duax et al., 1982), presence of hydrogen bond donor groups in the

17β side chain is responsible for glucocorticoid activity, whereas lack of these groups leads to

antiglucocorticoid activity of a compound. According to molecular docking calculations, amides

of cortienic acids with esterified L-amino acids might have glucocorticoid activity (Dobricic et

al., 2012). Binding energies of these compounds are lower than binding energies of

corresponding cortienic acids. Three-dimensional simulation of interaction between these

compounds and receptor for dexamethasone shows that carbonyl oxygen of L-amino acids forms

a hydrogen bond with ASN564. Additionally, this interaction is strengthen by hydrophobic

interactions between LEU563 and alkyl group used for esterification of the L-amino acid. These

��

�

interactions might be important for receptor activation and potential glucocorticoid activity of

these compounds. Interaction of amide of cortienic acid derived from hydrocortisone and methyl

ester of L-glycine with the receptor for dexamethasone is presented in Fig. 2.

< Fig. 2>

Twenty-two L-amino acid amides of cortienic acids obtained from hydrocortisone, prednisolone,

methylprednisolone, dexamethasone and betamethasone were synthesized and structurally

characterized. L-amino acids used for the synthesis are: methyl ester of L-glycine, methyl ester of

L-alanine, ethyl ester of L-glycine, ethyl ester of β-alanine and methyl ester of L-phenylalanine.

L-amino acids were chosen because they are natural constituents of human cells. As expected

metabolites of these compounds, they are good choice because they are not toxic. Synthesis of

these compounds is presented in Fig. 3.

< Fig. 3>

Compounds DG and DF (Fig. 3) had already been synthesized using N,N’-

dicyclohexylcarbodiimide (DCC) and HOBt according to Formstecher et al. (1980) and their

lypophilicity (Maes et al., 1988) as well as affinity for glucocorticoid receptor (Formstecher et

al., 1991) were determined. In this paper, alternative reaction employing EDC and HOBt (Anthes

et al., 2009) is used because of the ease of handling of EDC comparing to DCC (EDC is

crystalline powder, whereas DCC is low melting point waxy solid) and the enhanced solubility of

EDC and particularly the urea by-product formed during the reaction. Because of tert-amino

group in their structure, EDC and its urea by-product can be easily removed by extraction under

acidic conditions which facilitates purification. In addition, EDC is less sensitizing agent than

DCC.

https://www.researchgate.net/publication/244078616_High-performance_liquid_chromatography_in_the_evaluation_of_the_lipophilicity_of_17b-carboxamide_steroid_derivatives?el=1_x_8&enrichId=rgreq-90d0e066-b372-4b1f-afc3-7e1f72087969&enrichSource=Y292ZXJQYWdlOzI1ODYzNTEyNTtBUzoxMDI1NDAzNTYwOTYwMDFAMTQwMTQ1ODkzNDQ2Nw==

https://www.researchgate.net/publication/15823322_Synthesis_of_steroidal_17_beta-carboxamide_derivatives?el=1_x_8&enrichId=rgreq-90d0e066-b372-4b1f-afc3-7e1f72087969&enrichSource=Y292ZXJQYWdlOzI1ODYzNTEyNTtBUzoxMDI1NDAzNTYwOTYwMDFAMTQwMTQ1ODkzNDQ2Nw==

��

�

3.2. Physico-chemical characterization of synthesized compounds

�

Methyl 2-(11β,17α-dihydroxy-3-oxo-androst-4-en-17β-carboxamido) acetate (HG)

Light-gray crystalline solid obtained by purification with mobile phases C and D. Yield: 15.3%.

Melting point: 224-227.8 oC. IR (ATR) νmax (cm-1): 1237.91 (C-O ester), 1514.96 (N-H, bend),

1616.54 (C=C), 1654.84 (C=O amide), 1743.37 (C=O ester), 3424.01 (N-H, stretch). 1H NMR

(500 MHz, CD3OD) δ ppm 0.98 (3H, s, H-18), 1.00-1.02 (1H, m, H-9), 1.11 (1H, dq, J=9, J=13.5,

H-6), 1.40-1.44 (1H, m, H-15), 1.48 (3H, s, H-19), 1.51-1.56 (1H, m, H-16), 1.72 (1H, dd, J=2.7,

J=13.7, H-12), 1.77-1.78 (2H, m, H-14, H-15), 1.85-2.07 (4H, m, H-1, H-12, H-6, H-8), 2.20-

2.33 (3H, m, H-1, H-7, H-2), 2.46-2.60 (2H, m, H-2, H-7), 2.72-2.78 (1H, m, H-16), 3.73 (3H, s,

R-C(=O)OCH3), 3.95 (2H, ABq, J=17.5, R-NH-CH2-C(=O)OCH3), 4.40 (1H, q, J=3.2, H-11),

5.65 (1H, d, J=1, H-4). m/z = 418.3 (M+-1), 368.4, 386.2, 300.8. MS [M+H]+ calculated for

C23H33NO6 = 420.23079; observed = 420.23823. λmax(CH3OH) = 241 nm.

Methyl 2-(11β,17α−dihydroxy-3-oxo-androst-4-en-17β-carboxamido) propionate (HA)

White crystalline solid obtained by purification with mobile phases B and D. Yield: 17.4%.

Melting point: 122-125 oC. IR (ATR) νmax (cm-1): 1210.80 (C-O ester), 1512.32 (N-H, bend),


(500 MHz, CD3OD) δ ppm 0.96 (3H, s, H-18), 1.00 (1H, dd, J=3.5, J=11, H-9), 1.07-1.15 (1H,m,

H-6), 1.40-1.44 (1H, m, H-15), 1.40 (3H, d, J=7.5, R-NH-CH(CH3)-C(=O)OCH3), 1.48 (3H, s,

H-19), 1.49-1.54 (1H, m, H-16), 1.72-1.79 (3H, m, H-12, H-14, H-15), 1.85-1.96 (2H, m, H-1, H-

��

�

12), 2.04-2.07 (3H, m, H-1, H-7, H-2), 2.21-2.34 (3H, m, H-1, H-7, H-2), 2.46-2.60 (2H, m, H-2,

H-7), 2.72-2.77 (1H, m, H-16), 3.73 (3H, s, R-C(=O)OCH3), 4.39-4.44 (2H, m, H-11, R-NH-

CH(CH3)-C(=O)OCH3), 5.65 (1H, d, J=0.5, H-4). m/z = 432.0 (M+-1), 400.0, 382.1, 356.0. MS

[M+H]+ calculated for C24H35NO6 = 434.25371; observed = 434.25358. λmax(CH3OH) = 242 nm.

Ethyl 2-(11β,17α-dihydroxy-3-oxo-androst-4-en-17β-carboxamido) acetate (HEG)

White crystalline solid obtained by purification with mobile phases B and D. Yield: 32.2 %.



(500 MHz, CD3OD) δ ppm 0.99 (3H, s, H-18), 1.01 (1H, d, J=3.5, H-9), 1.12 (1H,dd, J=4, J=13,

H-6), 1.28 (3H, t, J=7, R-C(=O)OCH2CH3), 1.43-1.45 (1H, m, H-15), 1.48 (3H, s, H-19), 1.53-

1.56 (1H, m, H-16), 1.74-2.08 (7H, m, H-12, H-14, H-15, H-1, H-12, H-6, H-8), 2.22-2.33 (3H,

m, H-1, H-7, H-2), 2.46-2.57 (2H, m, H-2, H-7), 2.76 (1H, m, H-16), 3.94 (2H, ABq, J=17.5, R-

NH-CH2-C(=O)OCH2CH3), 4.20 (2H, q, J=7.2, R-C(=O)OCH2CH3), 4.40 (1H, d, J=3, H-11),

5.66 (1H, s, H-4). m/z = 432.3 (M+-1), 386.1, 368.2, 342.1. MS [M+H]+ calculated for

C24H35NO6 = 434.25372; observed = 434.25241. λmax(CH3OH) = 243 nm.

Ethyl 3-(11β,17α-dihydroxy-3-oxo-androst-4-en-17β-carboxamido) propionate (HEA)

White crystalline solid obtained by purification with mobile phases B and D. Yield: 31.9%.

Melting point: 178.7-181.7 oC. IR (ATR) νmax (cm-1): 1181.68 (C-O ester), 1518.98 (N-H, bend),

��

�


(500 MHz, CD3OD) δ ppm 0.96 (3H, s, H-18), 0.98 (1H,dd, J=3.5, J=11, H-9), 1.06-1.15 (1H, m,

H-6), 1.26 (3H, t, J=7.2, R-C(=O)OCH2CH3), 1.40-1.43 (1H, m, H-15), 1.47 (3H, s, H-19), 1.50-

1.57 (2H, m, H-16,H-12), 1.70-1.81 (2H, m, H-12, H-14, H-15), 1.85-1.93 (2H, m, H-1, H-12),

2.04-2.06 (2H, m, H-6, H-8), 2.20-2.33 (3H, m, H-1, H-7, H-12), 2.46-2.52 (1H, m, H-2), 2.55

(2H, t, J=6.7, R-NH-CH2CH2C(=O)OCH2CH3), 2.58-2.59 (1H, m, H-7), 2.73-2.78 (1H, m, H-

16), 3.42 (1H, dt, J=7, J=14, R-NH-CH2CH2C(=O)OCH2CH3), 3.51 (1H, dt, J=7, J=14, R-NH-

CH2CH2C(=O)OCH2CH3), 4.15 (1H, q, J=7.2, R-NH-CH2CH2C(=O)OCH2CH3), 4.38 (1H, q,

J=3.2, H-11), 5.65 (1H, d, J=0.5, H-4). m/z = 446.0 (M+-1), 346.1, 328.2, 301.1. MS [M+H]+

calculated for C25H37NO6 = 448.26937; observed = 448.27024. λmax(CH3OH) = 242 nm.

Methyl 2-(11β,17α-dihydroxy-3-oxo-androst-4-en-17β-carboxamido)-3-phenyl propionate (HF)

White crystalline solid obtained by purification with mobile phases A and D. Yield: 23.3%.


1612.05 (C=C), 1649.52 (C=O amide), 1664.19 (C3=O), 1742.95 (C=O ester), 3338.64 (N-H,

stretch). 1H NMR (500 MHz, CD3OD) δ ppm 0.93 (3H, s, H-18), 0.97 (1H, dd, J=3, J=11, H-9),

1.08-1.11 (1H, m, H-6), 1.39-1.42 (1H, m, H-15), 1.46 (3H, s, H-19), 1.51 (1H, m, H-16), 1.62

(1H, dd, J=2.5, J=14, H-12), 1.74 (2H, m, H-14, H-15), 1.87-2.05 (4H, m, H-1, H-12, H-6, H-8),

2.20-2.32 (3H, m, H-1, H-7, H-2), 2.45-2.69 (3H, m, H-2, H-7, H-16), 3.11 (2H, d, J=6.5, R-NH-

CH(CH2C6H5)-C(=O)OCH3), 3.70 (3H, s, R-NH-CH(CH2C6H5)-C(=O)OCH3), 4.67 (1H, t, J=6.5,

R-NH-CH(CH2C6H5)-C(=O)OCH3), 5.65 (1H, s, H-4), 7.19-7.31 (5H, m, R-NH-CH(CH2C6H5)-

��

�

C(=O)OCH3). m/z = 508.1 (M+-1), 476.2, 458.1, 174.0. MS [M+H]+ calculated for C30H39NO6 =

510.28502; observed = 510.28354. λmax(CH3OH) = 202 and 242 nm.

Methyl 2-(11β,17α-dihydroxy-3-oxo-androst-1,4-dien-17β-carboxamido) acetate (PG)

White crystalline solid obtained by purification with mobile phases C and E. Yield: 50%. Melting

point: 254.5-257 oC. IR (ATR) νmax (cm-1): 1203.08 (C-O ester), 1508.78 (N-H, bend), 1603.79

(C=C), 1650.59 (C=O amide), 1676.17 (C3=O), 1753.13 (C=O ester), 3385.72 (N-H, stretch). 1H

NMR (500 MHz, CD3OD) δ ppm 1.00 (3H, s, H-18), 1.02 (1H, d, J=4, H-9), 1.08-1.17 (1H, m,

H-6), 1.44-1.47 (1H, m, H-15), 1.50 (3H, s, H-19), 1.52-1.55 (1H, m, H-16), 1.70-1.77 (3H, m,

H-14, H-12, H-15), 1.93 (1H, dd, J=3.5, J=14, H-12), 2.14-2.18 (2H, m, H-6, H-8), 2.35-2.39

(1H, m, H-7), 2.62-2.69 (1H, m, H-7), 2.75 (1H, ddd, J=2.5, J=11, J=14, H-16), 3.73 (3H, s, R-

C(=O)OCH3), 3.95 (2H, ABq, J=17.5, R-NH-CH2-C(=O)OCH3), 4.40 (1H, q, J=3.5, H-11), 6.00

(1H, t, J=1.5, H-4), 6.24 (1H, dd, J=2, J=10, H-2), 7.48 (1H, d, J=10, H-1). m/z = 416.1 (M+-1),

366.2, 384.3, 340.2. MS [M+H]+ calculated for C23H31NO6 = 418.22241; observed = 418.22262.

λmax(CH3OH) = 244 nm.

Methyl 2-(11β,17α-dihydroxy-3-oxo-androst-1,4-dien-17β-carboxamido) propionate (PA)

White crystalline solid obtained by purification with mobile phases B and E. Yield: 30.5%.



��

�

(500 MHz, CD3OD) δ ppm 0.98 (3H, s, H-18), 1.01 (1H, dd, J=4, J=11, H-9), 1.08-1.17 (1H, m,

H-6), 1.40 (3H, d, J=7.5, R-NH-CH(CH3)-C(=O)OCH3), 1.43-1.54 (2H, m, H-15, H-16), 1.50

(3H, s, H-19), 1.67-1.80 (3H, m, H-14, H-12, H-15), 1.93 (1H, dd, J=3.5, J=14, H-12), 2.13-2.18

(2H, m, H-6, H-8), 2.35-2.39 (1H, m, H-7), 2.62-2.77 (2H, m, H-7, H-16), 3.74 (3H, s, R-

C(=O)OCH3), 4.39-4.44 (2H, m, H-11, R-NH-CH(CH3)-C(=O)OCH3), 6.00 (1H, t, J=1.7, H-4),

6.24 (1H, dd, J=2, J=10, H-2), 7.49 (1H, d, J=10, H-1). m/z = 430.1 (M+-1), 380.2, 398.2, 98.2.

MS [M+H]+ calculated for C24H33NO6 = 432.23806; observed = 432.23731. λmax(CH3OH) = 243

nm.

Ethyl 2-(11β,17α-dihydroxy-3-oxo-androst-1,4-dien-17β-carboxamido) acetate (PEG)

White crystalline solid obtained by purification with mobile phases B and E. Yield: 61%. Melting

point: 133.8-136.5 oC. IR (ATR) νmax (cm-1): 1209.26 (C-O ester), 1517.24 (N-H, bend), 1599.44

and 1613.71 (C=C), 1654.79 (C=O amide), 1675.99 (C3=O), 1728.90 (C=O ester), 3429.72 (N-H,

stretch). 1H NMR (500 MHz, CD3OD) δ ppm 1.01 (3H, s, H-18), 1.00-1.03 (1H, m, H-9), 1.09-

1.17 (1H, m, H-6), 1.28 (3H, t, J=7, R-C(=O)OCH2CH3), 1.44-1.48 (1H, m, H-15), 1.50 (3H, s,

H-19), 1.52-1.57 (1H, m, H-16), 1.70-1.79 (3H, m, H-14, H-12, H-15), 1.92-1.95 (1H, m, H-12),

2.14-2.18 (2H, m, H-6, H-8), 2.37 (1H, dd, J=3, J=13, H-7), 2.66 (1H, td, J=4.7, J=13.4, H-7),

2.75 (1H, ddd, J=2.5, J=11, J=14, H-16), 3.94 (2H, ABq, J=18, R-NH-CH2-C(=O)OCH2CH3),

4.20 (2H, q, J=7.2, R-C(=O)OCH2CH3), 4.40 (1H, q, J=3.2, H-11), 6.00 (1H, s, H-4), 6.24 (1H,

dd, J=2, J=10, H-2), 7.49 (1H, d, J=10, H-1). m/z = 430.0 (M+-1), 384.1, 366.1, 340.3. MS


��

�

Ethyl 3-(11β,17α-dihydroxy-3-oxo-androst-1,4-dien-17β-carboxamido) propionate (PEA)




(500 MHz, CD3OD) δ ppm 0.98 (3H, s, H-18), 0.99-1.01 (1H, m, H-9), 1.07-1.16 (1H, m, H-6),

1.26 (3H, t, J=7.2, R-C(=O)OCH2CH3), 1.43-1.48 (1H, m, H-15), 1.50 (3H, s, H-19), 1.50-1.56

(2H, m, H-16, H-12), 1.65-1.71 (1H, m, H-14), 1.74-1.80 (1H, m, H-15), 1.89 (1H, dd, J=4, J=14,

H-12), 2.15-2.17 (2H, m, H-6, H-8), 2.37 (1H, dd, J=3.2, 13.2, H-7), 2.55 (2H, t, J=6.7, R-NH-

CH2CH2C(=O)OCH2CH3), 2.62-2.68 (1H, m, H-7), 2.72-2.78 (1H, m, H-16), 3.42 (1H, dt, J=6.5,

13.5, R-NH-CH2CH2C(=O)OCH2CH3), 3.51 (1H, dt, J=6.5, 13.5, R-NH-

CH2CH2C(=O)OCH2CH3), 4.15 (2H, q, J=7.2, R-NH-CH2CH2C(=O)OCH2CH3), 4.38 (1H, q,

J=3.2, H-11), 6.00 (1H, s, H-4), 6.24 (1H, dd, J=2, J=10, H-2), 7.48 (1H, d, J=10, H-1). m/z =

444.0 (M+-1), 344.1, 326.1, 299.2. MS [M+H]+ calculated for C25H35NO6 = 446.25371; observed

= 446.25302. λmax(CH3OH) = 243 nm.

Ethyl 2-(11β,17α-dihydroxy-3-oxo-androst-1,4-dien-17β-carboxamido)-3-phenyl propionate (PF)

White crystalline solid obtained by purification with mobile phases A and E. Yield: 21.5%.


1601.03 and 1613.54 (C=C), 1653.80 (C=O amide), 1745.58 (C=O ester), 3320.67 (N-H,

stretch). 1H NMR (500 MHz, CD3OD) δ ppm 0.95 (3H, s, H-18), 0.98 (1H, dd, J=4, J=11, H-9),

1.10 (1H, ddd, J=4.5, J=13, J=17.5, H-6), 1.40-1.47 (1H, m, H-15), 1.48 (3H, s, H-19), 1.50-1.53

��

�

(1H, m, H-16), 1.72-1.78 (1H, m, H-15), 1.83 (1H, dd, J=3.5, J=14, H-12), 2.11-2.16 (2H, m, H-

6, H-8), 2.34 (1H, dd, J=3.5, J=13.5, H-7), 2.61-2.71 (2H, m, H-7, H-16), 3.10 (2H, d, J=6.7, R-

NH-CH(CH2C6H5)-C(=O)OCH3), 3.70 (3H, s, R-NH-CH(CH2C6H5)-C(=O)OCH3), 4.35 (1H, q,

J=3.2, H-11), 4.66 (1H, t, J=6.5, R-NH-CH(CH2C6H5)-C(=O)OCH3), 5.99 (1H, t, J=1.5, H-4),

6.23 (1H, dd, J=1.7, J=10.2, H-2), 7.18-7.29 (5H, m, R-NH-CH(CH2C6H5)-C(=O)OCH3), 7.46

(1H, d, J=10, H-1). m/z=506.3 (M+-1), 456.2, 364.9, 474.3. MS [M+H]+ calculated for

C30H37NO6 = 508.26936; observed = 508.26784. λmax(CH3OH) = 202 and 243 nm.

Methyl 2-(11β,17α-dihydroxy-6α-methyl-3-oxo-androst-1,4-dien-17β-carboxamido) acetate (MPG)

White crystalline solid obtained by purification with mobile phases C and E. Yield: 30%. Melting

point: 142.5-144.5 oC. IR (ATR) νmax (cm-1): 1210.68 (C-O ester), 1520.43 (N-H, bend), 1599.67

(C=C), 1651.32 (C=O amide), 1753.20 (C=O ester), 3350.68 (N-H, stretch). 1H NMR (500 MHz,

CD3OD) δ ppm 0.81 (1H, q, J=12.2, H-16), 0.97 (1H, dd, J=3.7, J=11.2, H-9), 1.01 (3H, s, H-18),

1.14 (3H, d, J=6.5, CH3 at C-6), 1.45-1.49 (1H, m, H-15), 1.50 (3H, s, H-19), 1.66-1.79 (3H, m,

H-14, H-12, H-15), 1.93 (1H, dd, J=3.7, J=14.2, H-12), 2.14-2.18 (1H, m, H-16), 2.24 (1H, ddd,

J=4, J=11, J=15, H-8) 2.73-2.78 (2H, m, H-7), 3.74 (3H, s, R-C(=O)OCH3), 3.95 (2H, ABq,

J=17.5, R-NH-CH2-C(=O)OCH3), 4.40 (1H, q, J=3.2, H-11), 5.99 (1H, t, J=1.5, H-4), 6.26 (1H,

dd, J=2, J=10, H-2), 7.50 (1H, d, J=10, H-1). m/z=430.1 (M+-1), 380.2, 398.2, 354.2. MS


Methyl 2-(11β,17α-dihydroxy-6α-methyl-3-oxo-androst-1,4-dien-17β-carboxamido) propionate

(MPA)

��

�


Melting point: 198.5-201 oC. IR (ATR) νmax (cm-1): 1210.97 (C-O ester), 1503.23 (N-H, bend),


stretch). 1H NMR (500 MHz, CD3OD) δ ppm 0.80 (1H, q, J=12.3, H-16), 0.96 (1H, d, J=3.5, H-

9), 0.99 (3H, s, H-18), 1.13 (3H, d, J=6.5, CH3 at C-6), 1.40 (3H, d, J=7.5, R-NH-CH(CH3)-

C(=O)OCH3), 1.44-1.48 (1H, m, H-15), 1.49 (3H, s, H-19), 1.51-1.54 (1H, m, H-6), 1.65-1.80

(3H, m, H-14, H-12, H-15), 1.92 (1H, dd, J=3.7, J=13.7, H-12), 2.13-2.17 (1H, m, H-16), 2.23

(1H, ddd, J=4.5, J=11.5, J=15.5, H-8), 2.72-2.77 (2H, m, H-7), 3.74 (3H, s, R-C(=O)OCH3),

4.39-4.44 (2H, m, H-11, R-NH-CH(CH3)-C(=O)OCH3), 5.99 (1H, t, J=1.5, H-4), 6.25 (1H, dd,

J=1.5, J=10, H-2), 7.49 (1H, d, J=10.5, H-1). m/z=444.0 (M+-1), 412.1, 394.0, 368.0. MS


Ethyl 2-(11β,17α-dihydroxy-6α-methyl-3-oxo-androst-1,4-dien-17β-carboxamido) acetate (MPEG)




(500 MHz, CD3OD) δ ppm 0.81 (1H, q, J=12.3, H-16), 0.97 (1H, dd, J=3.5, J=11.5, H-9), 1.01

(3H, s, H-18), 1.14 (1H, d, J=6.5, CH3 at C-6), 1.28 (3H, t, J=7.2, R-C(=O)OCH2CH3), 1.44-1.49

(1H, m, H-15), 1.50 (3H, s, H-19), 1.51-1.57 (1H, m, H-6), 1.66-1.79 (3H, m, H-14, H-12, H-15),

1.93 (1H, dd, J=3.7, J=13.7, H-12), 2.14-2.18 (1H, m, H-16), 2.24 (1H, ddd, J=4.5, J=11.5,

J=15.5, H-8), 2.73-2.78 (2H, m, H-7), 3.94 (2H, ABq, J=17.5, R-NH-CH2-C(=O)OCH2CH3),

��

�

4.20 (2H, q, J=7.2, R-C(=O)OCH2CH3), 4.40 (1H, q, J=3.2, H-11), 5.99 (1H, t, J=1.7, H-4), 6.26

(1H, dd, J=1.7, J=10.2, H-2), 7.50 (1H, d, J=10, H-1). m/z=444.1 (M+-1), 398.1, 380.1, 354.1.

MS [M+H]+ calculated for C25H37NO6 = 446.25372; observed = 446.25229. λmax(CH3OH) = 244

nm.

Ethyl 3-(11β,17α-dihydroxy-6α-methyl-3-oxo-androst-1,4-dien-17β-carboxamido) propionate

(MPEA)




(500 MHz, CD3OD) δ ppm 0.80 (1H, q, J=12.3, H-16), 0.96 (1H, dd, J=3.7, J=11.2, H-9), 0.98

(3H, s, H-18), 1.13 (1H, d, J=6.5, CH3 at C-6), 1.26 (3H, t, J=7, R-C(=O)OCH2CH3), 1.44-1.47

(1H, m, H-15), 1.49 (3H, s, H-19), 1.50-1.56 (2H, m, H-6,H-12), 1.63-1.69 (1H, m, H-14), 1.74-

1.77 (1H, m, H-15), 1.88 (1H, dd, J=3.7, J=13.7, H-12), 2.13-2.17 (1H, m, H-16), 2.22 (1H, ddd,

J=4, J=11, J=15, H-8), 2.54 (2H, t, J=6.7, R-NH-CH2CH2C(=O)OCH2CH3), 2.72-2.78 (2H, m, H-

7), 3.42 (1H, dt, J=7, J=14, R-NH-CH2CH2C(=O)OCH2CH3), 3.51 (1H, dt, J=6.5, J=13.5, R-NH-

CH2CH2C(=O)OCH2CH3), 4.15 (1H, q, J=7, R-NH-CH2CH2C(=O)OCH2CH3), 4.38 (1H, q,

J=3.3, H-11), 5.99 (1H, t, J=1.7, H-4), 6.25 (1H, dd, J=2, J=10, H-2), 7.49 (1H, d, J=10, H-1).

m/z=458.0 (M+-1), 358.1, 340.1, 440.7. MS [M+H]+ calculated for C26H37NO6 = 460.26937;

observed = 460.26874. λmax(CH3OH) = 243 nm.

��

�

Methyl 2-(11β,17α-dihydroxy-6α-methyl-3-oxo-androst-1,4-dien-17β-carboxamido)-3-phenyl

propionate (MPF)




stretch). 1H NMR (500 MHz, CD3OD) δ ppm 0.78 (1H, q, J=12.2, H-16), 0.93 (1H, d, J=3.5, H-

9), 0.95 (3H, s, H-18), 1.12 (1H, d, J=6.5, CH3 at C-6), 1.42-1.45 (1H, m, H-15), 1.48 (3H, s, H-

19), 1.49-1.53 (1H, m, H-6), 1.60-1.67 (2H, m, H-12, H-14), 1.74-1.76 (1H, m, H-15), 1.82 (1H,

dd, J=3.5, J=14, H-12), 2.11-2.16 (1H, m, H-16), 2.21 (1H, ddd, J=4, J=11, J=15, H-8), 2.66-2.75

(2H, m, H-7), 3.10 (2H, d, J=6.2, R-NH-CH(CH2C6H5)-C(=O)OCH3), 3.70 (3H, s, R-NH-

CH(CH2C6H5)-C(=O)OCH3), 4.35 (1H, q, J=3.2, H-11), 4.66 (1H, t, J=6.5, R-NH-

CH(CH2C6H5)-C(=O)OCH3), 5.98 (1H, t, J=1.5, H-4), 6.25 (1H, dd, J=2, J=10, H-2), 7.18-7.29

(5H, m, R-NH-CH(CH2C6H5)-C(=O)OCH3), 7.47 (1H, d, J=10, H-1). m/z=520.0 (M+-1), 470.2,

488.3, 174.1. MS [M+H]+ calculated for C31H39NO6 = 522.28502; observed = 522.28406.

λmax(CH3OH) = 202 and 243 nm.

Methyl 2-(11β,17α-dihydroxy-9α-fluoro-16α−methyl-3-oxo-androst-1,4-dien-17β-carboxamido)

acetate (DG)

White crystalline solid obtained by purification with mobile phases C and E. Yield: 17.2%.



��

�

stretch). 1H NMR (500 MHz, CD3OD) δ ppm 0.91 (1H, d, J=7.5, CH3 at C-16), 1.10 (3H, s, H-

18), 1.18-1.23 (1H, m, H-6), 1.53 (1H, ddd, J=5.5, J=13, J=18, H-15), 1.60 (3H, s, H-19), 1.64

(1H, d, J=2, H-12), 1.76 (1H, q, J=12, H-6), 1.86-1.91 (1H, m, H-15), 2.18-2.25 (2H, m, H-14, H-

12), 2.38-2.51 (2H, m, H-7, H-8), 2.69-2.76 (1H, m, H-7), 3.07-3.14 (1H, m, H-16), 3.74 (3H, s,

R-C(=O)OCH3), 3.96 (2H, ABq, J=17.5, R-NH-CH2-C(=O)OCH3), 4.25 (1H, dq, J=1.8, J=11,

H-11), 6.08 (1H, t, J=1.5, H-4), 6.29 (1H, dd, J=1.7, J=10.2, H-2), 7.43 (1H, d, J=10.5, H-1).

m/z=447.9 (M+-1), 396.1, 416.0, 311.0. MS [M+H]+ calculated for C24H32FNO6 = 450.22864;

observed = 450.22679. λmax(CH3OH) = 239 nm.

Methyl 2-(11β,17α-dihydroxy-9α-fluoro-16α-methyl-3-oxo-androst-1,4-dien-17β-carboxamido)

propionate (DA)

White crystalline solid obtained by purification with mobile phases B and E. Yield: 46%. Melting

point: 267.2-269.6 oC. IR (ATR) νmax (cm-1): 1505.40 (N-H, bend), 1606.06 and 1626.00 (C=C),

1645.36 (C3=O), 1665.51 (C=O amide), 1740.96 (C=O ester), 3398.25 (N-H, stretch). 1H NMR

(500 MHz, CD3OD) δ ppm 0.89 (1H, d, J=7, CH3 at C-16), 1.08 (3H, s, H-18), 1.18-1.23 (1H, m,

H-6), 1.41 (3H, d, J=7, R-NH-CH(CH3)-C(=O)OCH3), 1.53 (1H, ddd, J=5.5, J=13, J=18, H-15),

1.60 (3H, s, H-19), 1.64 (1H, dd, J=1.7, J=14.2, H-12), 1.75 (1H, q, J=12, H-6), 1.86-1.91 (1H,

m, H-15), 2.17-2.23 (2H, m, H-14, H-12), 2.38-2.51 (2H, m, H-7, H-8), 2.69-2.76 (1H, m, H-7),

3.08-3.13 (1H, m, H-16), 3.74 (3H, s, R-C(=O)OCH3), 4.25 (1H, dq, J=1.8, J=11, H-11), 4.42

(1H, q, J=7.2, R-NH-CH(CH3)-C(=O)OCH3), 6.08 (1H, s, H-4), 6.29 (1H, dd, J=2, J=10, H-2),

��

�

7.43 (1H, d, J=10, H-1). m/z=461.9 (M+-1), 430.1, 209.1, 410.1. MS [M+H]+ calculated for

C25H34FNO6 = 464.24429; observed = 464.24372. λmax(CH3OH) = 239 nm.

Ethyl 2-(11β,17α-dihydroxy-9α-fluoro-16α-methyl-3-oxo-androst-1,4-dien-17β-carboxamido)

acetate (DEG)



1622.17 (C=C), 1644.41 (C3=O), 1662.61 (C=O amide), 1735.59 (C=O ester), 3375.09 (N-H,


18), 1.18-1.23 (1H, m, H-6), 1.28 (3H, t, J=7, R-C(=O)OCH2CH3), 1.52 (1H, ddd, J=5.5, J=13,

J=18, H-15), 1.59 (3H, s, H-19), 1.62 (1H, dd, J=1.7, J=14.2, H-12), 1.76 (1H, q, J=12, H-6),

1.86-1.90 (1H, m, H-15), 2.17-2.25 (2H, m, H-14, H-12), 2.37-2.51 (2H, m, H-7, H-8), 2.69-2.75

(1H, m, H-7), 3.08-3.13 (1H, m, H-16), 3.94 (2H, ABq, J=17.5, R-NH-CH2-C(=O)OCH3), 4.19

(2H, q, J=7.2, R-NH-CH2-C(=O)OCH2CH3), 4.24 (1H, dq, J=2, J=11, H-11), 6.08 (1H, t, J=1.5,

H-4), 6.28 (1H, dd, J=1.7, J=10.2, H-2), 7.42 (1H, d, J=10.5, H-1). m/z=462.0 (M+-1), 396.2,

377.8, 416.1. MS [M+H]+ calculated for C25H34FNO6 = 464.24430; observed = 464.24286.

λmax(CH3OH) = 239 nm.

Ethyl 3-(11β,17α-dihydroxy-9α-fluoro-16α−methyl-3-oxo-androst-1,4-dien-17β-carboxamido)

propionate (DEA)

��

�





18), 1.17-1.22 (1H, m, H-6), 1.26 (3H, t, J=7, R-C(=O)OCH2CH3), 1.44 (1H, dd, J=1.5, J=14, H-

12), 1.50-1.53 (1H, m, H-15), 1.59 (3H, s, H-19), 1.75 (1H, q, J=11.3, H-6), 1.86-1.90 (1H, m, H-

15), 2.17-2.19 (2H, m, H-14, H-12), 2.37-2.50 (2H, m, H-7, H-8), 2.55 (2H, t, J=6.7, R-NH-

CH2CH2C(=O)OCH2CH3), 2.69-2.75 (1H, m, H-7), 3.08-3.13 (1H, m, H-16), 3.42 (1H, dt, J=7,

J=14, R-NH-CH2CH2C(=O)OCH2CH3), 3.53 (1H, dt, J=7, J=13.5, R-NH-

CH2CH2C(=O)OCH2CH3), 4.15 (2H, q, J=7.2, R-NH-CH2-C(=O)OCH2CH3), 4.21-4.24 (1H, m,

H-11), 6.08 (1H, s, H-4), 6.28 (1H, dd, J=1.7, J=10.2, H-2), 7.42 (1H, d, J=10.5, H-1). m/z=476.0

(M+-1), 376.1, 458.8, 311.1. MS [M+H]+ calculated for C26H36FNO6 = 478.25995; observed =

478.25876. λmax(CH3OH) = 240 nm.

Methyl 2-(11β,17α-dihydroxy-9α-fluoro-16α-methyl-3-oxo-androst-1,4-dien-17β-carboxamido)-3-

phenyl propionate (DF)




(500 MHz, CD3OD) δ ppm 0.83 (1H, d, J=7, CH3 at C-16), 1.03 (3H, s, H-18), 1.15-1.20 (1H, m,

H-6), 1.46-1.55 (2H, m, H-12, H-15), 1.57 (3H, s, H-19), 1.72 (1H, q, J=11.8, H-6), 1.84-1.88

��

�

(1H, m, H-15), 2.09-2.18 (2H, m, H-14, H-12), 2.36-2.48 (2H, m, H-7, H-8), 2.67-2.74 (1H, m,

H-7), 2.99-3.15 (3H, m, H-16, R-NH-CH(CH2C6H5)-C(=O)OCH3), 3.70 (3H, s, R-C(=O)OCH3),

4.19 (1H, dq, J=2, J=11, H-11), 4.68 (1H, t, J=6.7, R-NH-CH(CH2C6H5)-C(=O)OCH3), 6.07 (1H,

s, H-4), 6.28 (1H, dd, J=2, J=10, H-2), 7.19-7.29 (5H, m, R-NH-CH(CH2C6H5)-C(=O)OCH3),

7.40 (1H, d, J=10, H-1). m/z=538.3 (M+-1), 506.3, 173.9, 486.2. MS [M+H]+ calculated for

C31H38FNO6=540.27560; observed = 540.27405. λmax(CH3OH) = 240 nm.

Methyl 2-(11β,17α-dihydroxy-9α-fluoro-16β-methyl-3-oxo-androst-1,4-dien-17β-carboxamido)

acetate (BG)

White crystalline solid obtained by purification with mobile phases C and E. Yield: 42.9%.

Melting point: 263-266 oC. IR (ATR) νmax (cm-1): 1241.31 (C-O estar), 1503.75 (N-H, bend),


stretch). 1H NMR (500 MHz, CD3OD) δ ppm 1.14 (3H, s, H-18), 1.17-1.20 (1H, m, H-6), 1.22

(3H, d, J=7.5, CH3 at C-6), 1.52-1.58 (2H, m, H-12, H-15), 1.60 (3H, s, H-19), 1.94-2.14 (4H, m,

H-15, H-6, H-14, H-16), 2.19 (1H, dt, J=3.2, J=14, H-12), 2.41 (1H, dd, J=3.5, J=14, H-7), 2.46-

2.58 (1H, m, H-8), 2.71-2.78 (1H, m, H-7), 3.73 (3H, s, R-C(=O)OCH3), 3.93 (2H, ABq, J=17.5,

R-NH-CH2-C(=O)OCH3), 4.24 (1H, dq, J=2, J=10.5, H-11), 6.08 (1H, s, H-4), 6.28 (1H, dd, J=2,

J=10, H-2), 7.42 (1H, d, J=10, H-1). m/z=447.9 (M+-1), 398.0, 415.9, 396.0. MS [M+H]+

calculated for C24H32FNO6 = 450.22864; observed = 450.22826. λmax(CH3OH) = 239 nm.

��

�

Ethyl 2-(11β,17α-dihydroxy-9α-fluoro-16β-methyl-3-oxo-androst-1,4-dien-17β-carboxamido)

acetate (BEG)




stretch). 1H NMR (500 MHz, CD3OD) δ ppm 1.14 (3H, s, H-18), 1.18-1.20 (1H, m, H-6), 1.22

(3H, d, J=7.5, CH3 at C-6), 1.283 (3H, t, J=7, R-C(=O)OCH2CH3), 1.52-1.58 (2H, m, H-12, H-

15), 1.60 (3H, s, H-19), 1.93-2.14 (4H, m, H-15, H-6, H-14, H-16), 2.19 (1H, dt, J=3.4, J=14, H-

12), 2.41 (1H, dd, J=3.5, J=14, H-7), 2.46-2.58 (1H, m, H-8), 2.71-2.77 (1H, m, H-7), 3.92 (2H,

ABq, J=17.5, R-NH-CH2-C(=O)OCH3), 4.19 (2H, q, J=7.2, R-NH-CH2-C(=O)OCH2CH3), 4.24

(1H, dq, J=2, J=10.5, H-11), 6.08 (1H, s, H-4), 6.28 (1H, dd, J=2, J=10, H-2), 7.42 (1H, d, J=10,

H-1). m/z=461.9 (M+-1), 416.2, 395.9, 398.2. MS [M+H]+ calculated for C25H34FNO6 =

464.24430; observed = 464.24304. λmax(CH3OH) = 239 nm.

�

3.3. PAMPA test

The values of permeability coefficient–logPe, permeation parameter–CA(t)/CD(0), retention

factor–R and predicted human skin permeability–logKp of tested compounds are presented in

Table 2.

< Table 2 >

��

�

The results for hydrocortisone and dexamethasone obtained in this study were in good

accordance with previously reported data for these corticosteroids (Markovic et al., 2012).

According to Ottavaiani et al. (2006), compounds can be classified into 3 groups based on their

human skin permeability (logKp): compounds with lower logKp (logKp < -6) having negligible

membrane retention and low permeation (I) and compounds with higher logKp (logKp ≥ -6)

which have low or negligible membrane retention and high permeation (II) or high membrane

retention and low permeation (III). The proposed equation for correlation between logKp and

logPe is:

logKp = (1.34 ± 0.12) logPe + (0.28 ± 0.56) (3)

Using this equation, logKp values for all tested compounds were calculated (Table 1). All tested

compounds have logKp < -6 and belong to group I (compounds with negligible membrane

retention and low permeation) (Fig. 4).

< Fig. 4 >

The majority of 17β−carboxamide derivatives have retention similar to the retention of starting

corticosteroids. The next step was to identify derivatives that have significantly higher retention

than starting corticosteroids. Higher retention of a compound can favorably affect its potential

anti-inflammatory activity because the compound is retained in the site of action in higher

amount. On the other hand, higher permeability can cause systemic side effects and reduce local

activity of a compound. According to the structure and expected metabolic properties of these

compounds, low systemic toxicity is expected because these compounds should be chemically

transformed to inactive and non-toxic metabolites. Therefore, higher permeability should

��

�

negatively affect only potential local anti-inflammatory activity of these compounds. Single

factor ANOVA was used to test the triplicates of results for each compound, comparing their

mean values in order to identify derivatives with significantly higher retention than

corticosteroids from which they were synthesized. Significantly higher retention was obtained for

MPEA and all derivatives which are amides of cortienic acid and methyl ester of L-phenylalanine

(HF, PF, MPF, DF). Apart from this, phenylalanine derivatives have significantly higher

permeability than starting corticosteroids. Derivative MPEA doesn’t have higher permeability

than methylprednisolone which distinguishes this compound from phenylalnine derivatives.

3.4. Stepwise MLR-, PLS-, and ANN-QSPR studies

The quantitative structure-property relationship (QSPR) study was performed to investigate the

correlations between logPe and R (dependent variables) of the examined compounds and their

calculated constitutional, geometrical, thermodynamic, physicochemical and electronical

molecular parameters (independent variables). Independent variables that have the strongest

influence on dependent variables (logPe and R) and form final QSPR models (logPe and R) are

defined as molecular descriptors. In order to perform relevant comparison between different

methodologies used to build QSPR models, same training, verification and test sets were used.

For MLR(logPe) and PLS(logPe) test set consisted of 11 compounds (BEG, HA, MPF, PEG, PG,

HEA, HEG, HG, MPA, MPEG, MPG) while other compounds were chosen as training set. For

ANN(logPe), compounds used as test set for MLR(logPe) and PLS(logPe) were divided into two

groups: verification set (HEA, HEG, HG, MPA, MPEG, MPG - 6 compounds) and test set

(BEG, HA, MPF, PEG, PG - 5 compounds). For MLR(R) and PLS(R) test set consisted of 11

��

�

compounds (HEA, HEG, HG, MPA, MPEG, PG, BEG, HA, MPF, MPG, PEG), while other

compounds were chosen as training set. For ANN(R), compounds used as test set for MLR(R)

and PLS(R) were divided into two groups: verification set (HEA, HEG, HG, MPA, MPEG, PG -

6 compounds) and test set (BEG, HA, MPF, MPG , PEG - 5 compounds). Test and verification

sets were formed in the way that logPe and R values of these compounds were homogenously

distributed in the whole range of logPe and R values.

Multilayer perceptron artificial neural network based on backpropagation training algorithm was

chosen for nonlinear QSPR modeling. The first step in ANN construction was appropriate

independent variable selection. The genetic algorithm (Gupta et al., 2011), principal component

analysis (Zhang, 2007), and stepwise MLR (Filipic et al., 2013; Gonzalez-Arjona et al., 2002;

Jalali-Heravi and Garkani-Nejad, 2002) had been used for independent variable selection. In this

study, stepwise MLR was applied. Thus selected independent variables were used as inputs for

ANN modeling. With forward stepwise MLR, independent variables are added one by one into

the model. These variables are evaluated at each step, being added or deleted from the model

based on specified criteria (F to enter and F to remove criteria). For ANN(logPe), F to enter

values were 7 and 8, whereas F to remove value in both cases was 6. For ANN(R), F to enter

values were 4 and 6, whereas F to remove value was 3. Using stepwise MLR, sets of 7 and 4

independent variables for both ANN(logPe) and ANN(R) were formed. Subsequently, three-

layer networks were created. The first layer is input layer. Number of nodes in the input layer is

equivalent to the number of selected independent variables. The second layer is a hidden layer

with optimal number of units (nodes). The third layer is output layer, consisting of one node

(logPe or R of the examined compounds). The whole data set was divided into training (16

compounds), verification (6 compounds) and test set (5 compounds). The training set is used for

https://www.researchgate.net/publication/49673271_Prediction_of_capillary_gas_chromatographic_retention_times_of_fatty_acid_methyl_esters_in_human_blood_using_MLR_PLS_and_back-propagation_artificial_neural_networks?el=1_x_8&enrichId=rgreq-90d0e066-b372-4b1f-afc3-7e1f72087969&enrichSource=Y292ZXJQYWdlOzI1ODYzNTEyNTtBUzoxMDI1NDAzNTYwOTYwMDFAMTQwMTQ1ODkzNDQ2Nw==

https://www.researchgate.net/publication/233538966_Quantitative_structure-mobility_relationship_analysis_of_imidazoline_receptor_ligands_in_CDs-mediated_CE?el=1_x_8&enrichId=rgreq-90d0e066-b372-4b1f-afc3-7e1f72087969&enrichSource=Y292ZXJQYWdlOzI1ODYzNTEyNTtBUzoxMDI1NDAzNTYwOTYwMDFAMTQwMTQ1ODkzNDQ2Nw==

��

�

network training and the verification set is used to perform an independent check of the network

performance during training in order to avoid overfitting the data. Finally, the test set is used to

provide a final independent check of the network performance. The optimal number of nodes in

hidden layer, number of epochs, momentum and learning rate for both ANN(logPe) and ANN(R)

modeling were previously determined in initial network trainings for each set of chosen

descriptors for ANN(logPe) and ANN(R), evaluating prediction errors for training, verification

and test sets. Thus formed models with different number of descriptors (7 and 4 for both

ANN(logPe) and ANN(R)) were analyzed and best ones were selected comparing Q2 (cross-

validated squared correlation coefficient), r (correlation between observed and predicted values

of test set), RMSEE (root mean squared error of estimation) and RMSEP (root mean squared

error of prediction) values. The final structure of selected models (number of nodes in input layer

- number of nodes in hidden layer - number of nodes in output layer) was 4-6-1 (ANN(logPe))

and 4-4-1 (ANN(R)).

Forward stepwise MLR models have been developed to assess the linear relationship between

calculated molecular parameters and logPe and R of tested compounds. The sets of 4 independent

variables used for ANN(logPe) and ANN(R) were employed for MLR modeling. Using forward

stepwise method (the same F to enter and F to remove values and other parameters as for the

independent variable selection for ANN modeling), three out of four descriptors from the selected

ANN(logPe) model were included in the final MLR(logPe) model, whereas entire set of four

descriptors from the selected ANN(R) model was included in the final MLR(R) model (Table 3).

In contrast to MLR analysis, PLS approach can analyze data with strongly collinear, noisy, and

numerous independent variables (X-variables). In PLS modeling, the importance of each X-

variable is evaluated according to its VIP value. The X-variables with VIP value larger than 1 are

��

�

the most relevant, those with 1.0>VIP>0.5 are moderately influential, while X-variables with

VIP value smaller than 0.5 are not relevant for the model (Eriksson et al., 2001; Umetrics AB,

2008). X-variables with lowest VIP-values are successively removed from the PLS model and

each time new PLS model is created. For each created model regression factors R2, Q2, F ratio, p

value, RMSEE are calculated and compared with the previous model. The procedure is repeated

until the best model is obtained. The response permutation test (Y scrambling) is used to examine

the statistical significance of the R2 and Q2 and overfitting due to the chance correlation (Eriksson

et al., 2001). In the response permutation test, the Y-matrix is randomly re-ordered (100 times in

this project) whereas the X-matrix is kept intact. Model is fitted to the new Y-data and the new

R2, Q2 and VIP parameters are calculated. All model selection steps are repeated on the

scrambled Y-response data. Lines are fitted through the R2-values and through the Q2-values to

yield two separate intercepts. For a valid model, the R2-intercept should not exceed 0.4 while the

Q2-intercept should be lower than 0.05 (Eriksson et al., 2001).

Statistical parameters calculated for QSPR models are presented in Table 3.

< Table 3>

Numerical values of descriptors that form final QSPR models are presented in Supplementary

tables 2 and 3.

The quality of the obtained QSPR models was estimated using such parameters, as R2 (square of

the correlation coefficient), the F ratio, the p value, RMSEE, RMSEP, r, Q2 (equation (5)) and

R2pred (equation (6)).

RMSEE value is calculated for training set, whereas RMSEP value is calculated for test and

verification set (ANN models) or test set (MLR and PLS models).

��

�

Q2 is an internal validation parameter. It is calculated using leave-one-out (LOO) cross-validation

technique where each compound of the training set is deleted once while the remaining

compounds are used to create a model. The value of the deleted compound is predicted using the

developed model. The procedure is repeated until all the compounds are deleted once (Ojha et

Roy, 2011; Snedecor et Cochran, 1967). Q2 is calculated according to the equation (5). In this

equation, Y training is average logPe or R value of compounds from training set, while Yobs(training) is

an observed logPe or R value from the training set. PRESS is calculated for the training set

according to the equation (4) after the completion of the LOO procedure. QSPR models with Q2≥

0.5 are considered to have good predictive capability (Eriksson et al., 2001; Umetrics AB, 2008).

Q2 value enables determination of predictive potential of a QSPR model for compounds that are

similar to ones in the training set. To determine predictive potential of a QSPR model for

compounds that differ in a certain manner from those in the training set, R2pred is used. R2

pred is

an external validation parameter, calculated according to the equation (6) (Marshall, 1994). In

equation (6), Yobs(test) is an observed value of logPe or R of a test set compound and Y training is

mean logPe or R value of the training set compounds. PRESS is calculated for test set according

to the equation (4).

∑ == n

i iePRESS1

2)( ��(4)�

Difference between observed and the predicted Y values - (e(i))

�� ∑ −−=

2)(

2

)(1Q

trainingtrainingobs YY

PRESS

��(5)

��

∑ −−=

2)(

2

)(1R

trainingtestobs

pred

YY

PRESS��(6)�

�

��

�

QSPR models with R2pred≥ 0.5 can be considered to have good predictive capability (Ojha et Roy,

2011; Tropsha, 2010).

The F-test, which is based on the ratio MS Regression/MS Residual, formally evaluates

significance of the model. The p value indicates the probability level where a model with this F

value may be the result of just chance. Models with p-value lower than 0.05 are considered

significant (Eriksson et al., 2001).

Analyzing Q2 and R2pred for obtained QSPR models (Table 4), it can be concluded that all of them

have good predictive potential (Q2≥0.5 and R2pred≥ 0.5). Comparing other statistical parameters,

the most reliable QSPR models for logPe and R were PLS(logPe) and ANN(R).

3.4.1. Interpretation of selected QSPR models

According to VIP parameters of descriptors in selected PLS(logPe) model, all of them have

similar influence on the permeability of tested compounds through the artificial PAMPA

membrane. VIP values of these descriptors are from 0.9 to 1.2 (Fig. 5).

< Fig. 5 >

Coefficient plot (Fig. 6) shows that all descriptors have positive, apart from D/Dtr06 which has

negative influence on logPe. Descriptor D/Dtr06 is a ring descriptor and represents

distance/detour ring index of order 6. Descriptors F02[C-C] and F05[C-O] represent frequency of

C-C at topological distance 2 and frequency of C-O at topological distance 5. These descriptors

belong to 2D atom pairs descriptors. MLOGP2 is a squared Moriguchi octanol-water partition

coefficient. CATS2D_08_LL belongs to CATS2D descriptors. This descriptor is defined as

CATS2D lipophilic-lipophilic at lag08 (TALETE srl, 2013). MLOGP2 and CATS2D_08_LL are

��

�

related to the lipophilicity of tested compounds. Apart from CATS2D_08_LL, CATS2D_09_DL

appears in other two models (ANN(logPe) and MLR(logPe)) and also represents lipophilicity.

This indicates that lipophilicity of tested molecules is important property that influences

permeability through the artificial PAMPA membrane.

< Fig. 6>

Analyzing response graphs obtained from the ANN(R) model, which show the influence of each

descriptor on retention of tested compounds, it can be concluded that nBM, Mor24s and

RDF145s have positive, whereas RDF015m has negative influence on retention (Fig. 7). The

same conclusion can be drawn analyzing MLR(R) model and coefficients in front of each

descriptor.

< Fig. 7>�

According to the sensitivity analysis of the descriptors in ANN(R) model (Table 4), the most

influential descriptor on retention of tested compounds in the PAMPA artificial membrane is

nBM (number of multiple bonds) (TALETE srl, 2013). The increase of the value of these

descriptors causes the increase of retention. This can be used to explain why the majority of

prednisolone derivatives (PG, PEG, PA, PF) have higher retention than corresponding derivatives

of hydrocortisone (HG, HEG, HA, HF). In addition, this can be used to explain high retention of

phenylalanine derivatives (HF, PF, MPF, DF). According to this, π−π interaction between

molecules and constituents of the artificial membrane might be one of membrane retention

mechanisms of these compounds.

< Table 4 >

��

�

Mor24s is a 3D MoRSE descriptor (TALETE srl, 2013). The 3D-MoRSE descriptors are

molecule atom projections along different angles, such as in electron diffraction. These

descriptors take into account 3D arrangement of the atoms in a molecule without ambiguities. 3D

MoRSE descriptors do not depend on the molecular size and can be applied to a large number of

molecules with great structural variance (Duchowicz et al., 2006). They represent different views

of the whole structure, but their meaning is not clear (Todeschini and Consonni, 2000; Hancock

et al., 2005). RDF015m and RDF145s are Radial Distribution Function (RDF) descriptors

weighted by mass and l-state respectively (TALETE srl, 2013). RDF015m represents 3D

molecular distribution of mass calculated at radius of 1.5 Å from the geometrical center of a

molecule. This descriptor is actually related to steric factors at the radius of 1.5 Å from the

geometrical center of the molecule (Gonzalez et al., 2006). Compounds with the highest values of

this descriptor are DF, DEG, DG and DEA. Most likely, this is related to the presence of methyl

group in α-orientation at position C16 in these compounds, since the distance between this group

and geometrical centers of these molecules is 2-2.5 Å. This is presented in Fig. 8 (DG is used as

an example, where geometrical center (Du) and methyl group at C16 are circled). RDF145s is a

signal calculated at radius of 14.5 Å from the center of molecule. BEG and MPEA have highest,

whereas prednisolone and methylprednisolone have lowest values of this descriptor. This

descriptor could be used to explain positive influence of side chain length on the retention.

< Fig. 8 >

Instead of PAMPA experiments, formed QSPR models could be used for in silico prediction of

human skin permeability and retention of novel 17β−carboxamide steroids (e.g.,obtained by

structural modifications of compounds tested in this study or 17β−carboxamide steroids derived

from other corticosteroids).

��

�

4. Conclusion

Twenty-two 17β-carboxamide derivatives of hydrocortisone, prednisolone, methylprednisolone,

dexamethasone and betamethasone were synthesized and structurally characterized. These

compounds are L-amino acid amides of cortienic acids obtained by periodic acid oxidation of

corresponding corticosteroids and represent potential soft corticosteroids with local anti-

inflammatory activity in the skin. Their permeability and retention in the human skin were

estimated using PAMPA test and permeability coefficients, permeation parameters, retention

factors and predicted human skin permeability values were calculated. According to calculated

permeation parameters and retention factors, all compounds are classified into group I

(compounds with negligible membrane retention and low permeation). Compounds with

significantly higher retention than corresponding corticosteroids were identified. High retention

of a compound in the skin can positively affect its potential local anti-inflammatory activity with

minimal systemic side effects. Derivatives HF, PF, MPF, DF and MPEA have significantly

higher retention than corresponding corticosteroids while compound MPEA doesn’t have

significantly higher permeability than methylprednisolone. Geometrical, thermodynamic,

physicochemical and electronical molecular parameters were calculated and stepwise MLR-,

PLS-, and ANN-QSPR studies were performed. Best models that explain permeability

coefficients and retention in the artificial membrane were chosen. Analyzing descriptors in these

models, the most significant molecular properties that influence permeability and retention in the

artificial membrane were identified. Taking this into account, further structural modifications of

these compounds could be applied in order to improve their retention without significant increase

��

�

of permeability. High retention and low permeability, together with expected metabolical

inactivation of 17β-carboxamide steroids, should enable favourable therapeutical effects with

minimal local and systemic side effects, which are common for traditional corticosteroids when

applied to the skin. Selected QSPR models provide in silico prediction of human skin

permeability and retention of novel 17β−carboxamide steroids without performing PAMPA

experiments.

Acknowledgements

�

This work was financially supported by the Ministry of Education, Science and Technological

Development, Belgrade, Serbia, as part of Project No.172041. We also thank the Institute of

Chemistry, Technology and Metalurgy, Belgrade, Serbia and Department of Biochemistry,

Faculty of Chemistry, University of Belgrade, Serbia for their assistance.

References

Anthes, J. C., McCormick, K.D., Hey, J., A., Aslanian, R. G., Biju, P. J., Berlin, M. Y., Solomon,

D. M., Wang, H., Lim, Y.-H., Lee, Y.J., Bitar, R., D., 2009. C20-C21 Substituted

glucocorticoid receptor agonists. WO 2009/085879 A2.

Avdeef, A., Strafford, M., Block, E., Balogh, M., Chambliss, W., Khan, I., 2001. Drug absorption

in vitro model: filter-immobilized artificial membranes. Studies of the permeability

properties of lactones in Piper methysticum Forst. Eur. J. Pharm. Sci. 14, 271-280.

��

�

Avdeef, A., 2003. Absorption and Drug Development, John Wiley & Sons: Hoboken, New

Jersey.

Bladh, H., Burkamp, F., Gabos, B., Hansen, P., 2010. Novel derivatives of steroidal[3,2-

c]pyrazole compounds with glucocorticoid activity. WO 2010/114476 A1.

Bodor, N., Bodor, N., Wu, WM., 1992a. A comparison of intraocular pressure elevating activity

of loteprednol etabonate and dexamethasone in rabbits. Curr. Eye. Res. 11, 525–530.

Bodor, N., Loftsson, T., Wu, WM., 1992b. Metabolism, distribution, and transdermal

permeability of a soft corticosteroid, loteprednol etabonate. Pharm. Res. 9, 1275–1278.

Bodor, N., Buchwald, P., 2006. Corticosteroid design for the treatment of asthma: structural

insights and the therapeutic potential of soft corticosteroids. Curr. Pharm. Des. 12, 3241-

3260.

CambridgeSoft Corporation, 2005. Chem3D Ultra, version 9.0.1. Cambridge, MA, USA.

ChemAxon, 2008. MarvinSketch 5.1.3. Budapest, Hungary.

Dobricic, V., Markovic, B., Savic, V., Milenkovic, N., Vladimirov, S., Cudina. O., 2012.

Molecular docking studies, synthesis and structural characterization of two novel soft

corticosteroids. 11th International Conference on Fundamental and Applied Aspects of

Physical Chemistry, Belgrade, Serbia, September 24-28, 2012, pp 576-578.

Druzgala, P., Hochhaus, G., Bodor, N., 1991a. Soft drugs-10. Blanching activity and receptor

binding affinity of a new type of glucocorticoid: Loteprednol etabonate. J Steroid

Biochem. Mol. Biol. 38, 149–154.

Druzgala, P., Wu, WM., Bodor, N., 1991b. Ocular absorption and distribution of loteprednol

etabonate, a soft steroid, in rabbit eyes. Curr. Eye. Res. 10, 933–937.

��

�

Duax, W. L., Griffin, J. F., Rohrer, D. C., Wecks, C. M., 1982. In: Hormone Antagonists

(Agarwal, M. K., ed.), Walter de Gruyter, Berlin.

Duchowicz, P.R., Fernandez, M., Caballero, J., Castro, E.A, Fernandez, F.M., 2006. QSAR for

non-nucleoside inhibitors of HIV-1 reverse transcriptase. Bioorg. Med. Chem. 14, 5876–

5889.

Eriksson, L., Johansson, E., Kettaneh-Wold, N., Trygg, J., Wikstrom, C., Wold, S., 2001. Multi-

and Megavariate Data Analysis. Basic Principles and Applications I, 2nd ed., Umetrics

Academy, Umeå.

Filipic, S., Nikolic, K., Vovk, I., Krizman, M., Agbaba, D., 2013. Quantitative structure-mobility

relationship analysis of imidazoline receptor ligands in CDs-mediated CE.

Electrophoresis 34, 471–482.

Formstecher, P., Lustenberger, P., Dautrevaux, M., 1980. Synthesis of steroidal 17β-carboxamide

derivatives. Steroids 35, 265-272.

Formstecher, P.A., Lefebvre, Ph., Burollaud, T., 1991. Hormones and antihormones. The

steroidal model. J. Pharm. Belg. 46, 37-48.

Gonzalez-Arjona, D., Lopez-Perez, G., Gustavo-Gonzalez, A., 2002. Non-linear QSAR modeling

by using multilayer perceptron feedforward neural networks trained by back-propagation.

Talanta 56, 79–90.

Gonzalez, M.P., Puente, M., Fall, Y., Gomez, G., 2006. In silico studies using Radial Distribution

Function approach for predicting affinity of 1α,25-dihydroxyvitamin D3 analogues for

Vitamin D receptor. Steroids 71, 510–527.

Gupta, V. K., Khani, H., Ahmadi-Roudi, B., Mirakhorli, S., Fereyduni, E., Agarwal, S., 2011.

Prediction of capillary gas chromatographic retention times of fatty acid methyl esters in

��

�

human blood using MLR, PLS and back-propagation artificial neural networks. Talanta

83, 1014–1022.

Hancock, T., Put, R., Coomans, D., Heyden, Y.V., Everingham, Y., 2005. A performance

comparison of modern statistical techniques for molecular descriptor selection and

retention prediction in chromatographic QSRR studies. Chemometr. Intell. Lab. 76, 185–

196.

Iczkowski, R. P., Margrave, J. L., 1961. Electronegativity. J. Am. Chem. Soc. 83, 3547–3551.

Jalali-Heravi, M., Garkani-Nejad, Z., 2002. Prediction of electrophoretic mobilities of alkyl- and

alkenylpyridines in capillary electrophoresis using artificial neural networks. J.

Chromatogr. A 971, 207–215.

Kansy, M., Senner, F., Gubernator, K., 1998. Physicochemical high-throughput screening:

Parallel artificial membrane permeation assay in the description of passive absorption

processes. J. Med. Chem 41, 1007–1010.

Maes, P., Formstecher, P., Lustenberger, P., Dautrevaux, M., 1988. High-performance liquid

chromatography in the evaluation of the lipophilicity of 17β-carboxamide steroid

derivatives. J. Chromatogr. 445, 409-416.

Manz, B., Grill, J., Kreienberg, R., Pollow, K., 1983. Methyl 17β-carboxyester derivatives of

natural and synthetic glucocorticoids: correlation between receptor binding and inhibition

of in vitro phytohaemagglutinin-induced lymphocyte blastogenesis. J. Clin. Chem. Clin.

Biochem. 21, 69-75.

Manz, B., Rehder, M., Heubner, A., Kreienberg, R., Grill, H.J., Pollow, K., 1984. 17β-

carboxamide steroids: highly effective inhibitors of the phytohaemagglutinin mediated

��

�

blastogenesis of normal human peripheral lymphocytes. J. Clin. Chem. Clin. Biochem.

22, 209-214.

Markovic, B., Vladimirov, S., Cudina, O., Odovic, J., Karljikovic-Rajic, K., 2012. A PAMPA

assay as fast predictive model of passive human skin permeability of new synthesized

corticosteroid C-21 esters. Molecules 17, 480-491.

Marshall, G.R., 1994. Binding-site modeling of unknown receptors, in: H. Kubinyi (Ed.), 3D

QSAR in Drug Design – Theory, Methods and Applications. ESCOM, Leiden, pp. 80–

116.

Microsoft Corporation, 2010. Microsoft Office Excel 2010. Redmond, WA, USA.

Ojha, P.K., Roy, K., 2011. Comparative QSARs for antimalarial endochins: Importance of

descriptor-thinning and noise reduction prior to feature selection. Chemometr. Intell. Lab.

109, 146–161.

Ottaviani, G., Martel, S., Carrupt, P.-A., 2006. Parallel artificial membrane permeability assay: a

new membrane for the fast prediction of passive human skin permeability. J. Med. Chem.

49, 3948-3954.

Parr, R. G., Yang, W., 1989. Density-Functional Theory of Atoms and Molecules, Oxford

University Press, New York.

Rousseau, G., Kirchhoff, J., Formstecher, P., Lustenberger, P., 1979. 17β-carboxamide steroids

are a new class of glucocorticoid antagonists. Nature 279, 158-160.

Sinkó, B., Kökösi, J., Avdeef, A., Takács-Novák, K., 2009. A PAMPA study of the permeability-

enhancing effect of a new ceramide analogues. Chem. Biodivers. 6, 1867–1874.

��

�

Sinkó, B., Garrigues, T.M., Balogh, G.T., Zsombor, K.N., Tsinman, O., Avdeef, A., Takács-

Novák, K., 2012. Skin–PAMPA: A new method for fast prediction of skin penetration.

Eur J Pharm Sci 45, 698–707.

Snedecor, G.W., Cochran, W.G., 1967. Statistical Methods, Oxford & IBH, New Delhi.

StatSoft, Inc., 1998. STATISTICA neural networks 4.0. Tulsa, OK, USA.

TALETE srl, 2010. Dragon 6.0. Milano, Italy.

TALETE srl, 2013. Molecular descriptors, QSAR, chemometrics and chemoinformatics-dragon

molecular descriptor list.

<www.talete.mi.it/products/dragon_molecular_descriptor_list.pdf>.

Todeschini, R., Consonni, V., 2000. Handbook of Molecular Descriptors, Wiley-VCH,

Weinheim.

Tropsha, 2010. Best practices for QSAR model development, validation and exploration. Mol.

Inf. 29, 476-488.

Umetrics AB, 2008. SIMCA P+ program, Version 12.0.0.0. Umea, Sweden.

Zhang, Y. X., 2007. Artificial neural networks based on principal component analysis input

selection for clinical pattern recognition analysis. Talanta 73, 68–75.

��

�

Figure captions

Fig. 1. Periodic acid oxidation of corticosteroids.

Fig. 2. Three-dimensional simulation of interaction between selected compound and receptor for

dexamethasone.

Fig. 3. Synthesis of amides.

Fig. 4. The relationship between retention factor, R and permeation parameter, CA(t)/CD(0) of

examined compounds in PAMPA test.

Fig. 5. VIP values of descriptors in selected PLS(logPe) model.

Fig. 6. Coefficient plot of descriptors in selected PLS(logPe) model.

Fig. 7. Response graphs from the ANN(R) model.

Fig. 8. 3D structure with geometrical center of DG.

��

�

Table 1

Molecular parameters calculated in MarvinSketch and Chem3D Ultra 9.0.1.

Programs Molecular parameters

MarvinSketch Platt index, Randic index, Harary index, hyper wiener index, Szeged index, Wiener index,

Wiener polarity, dreiding energy, minimal projection area, maximal projection area,

minimal projection radius, maximal projection radius, polar surface area, molecular surface

area

Chem3D Ultra 9.0.1 dipol, charge at C1 atom, charge at C2 atom, electronic density at C1 atom, electronic

density at C2 atom, highest occupied molecular orbital (HOMO) energy, lowest unoccupied

molecular orbital (LUMO) energy, chemical potential, electronegativity, hardness, global

softness, steric energy summary, Connoly accessible area, Connoly molecular area, Connoly

solvent excluded volume, molecular weight, ovality, Balaban index, cluster count, diameter,

radius, molecular topological index, shape attribute, shape coefficient, sum of degrees, sum

of valence degrees, total connectivity, total valence connectivity, Wiener index, molar

refractivity, partition coefficient, bend energy, non-1,4-VDW energy, stretch energy, stretch

bend energy, torsion energy, total energy, VDW 1,4 energy, electronic energy

�

��

�

Table 2

Calculated PAMPA parameters of tested compounds.

Compound CA(t)/CD(0) logPe R logKp

Hydrocortisone 0.72±0.03 -6.50±0.02 1.60±0.17 -8.44±0.02

HG 0.31±0.00 -6.88±0.01 0.33±0.14 -8.94±0.01

HA 0.89±0.13 -6.42±0.06 0.34±0.23 -8.32±0.09

HEG 0.47±0.04 -6.70±0.04 0.55±0.33 -8.70±0.05

HEA 1.01±0.09 -6.36±0.04 1.98±0.53 -8.24±0.05

HF 10.99±0.48 -5.23±0.00 8.42±3.59 -6.73±0.01

Prednisolone 0.67±0.02 -6.53±0.02 1.26±0.64 -8.48±0.02

PG 0.17±0.01 -7.13±0.03 0.75±0.43 -9.27±0.04

PA 0.55±0.08 -6.62±0.06 3.28±1.21 -8.59±0.09

PEG 0.49±0.03 -6.67±0.03 2.15±0.49 -8.65±0.04

PEA 0.63±0.04 -6.57±0.03 0.67±0.12 -8.52±0.04

PF 8.07±0.81 -5.37±0.05 11.01±0.84 -6.92±0.06

Methylprednisolone 1.35±0.06 -6.23±0.02 0.94±0.17 -8.19±0.03

MPG 0.55±0.05 -6.62±0.04 1.23±0.31 -8.59±0.06

MPA 1.37±0.06 -6.22±0.02 1.67±0.12 -8.06±0.03

MPEG 0.94±0.16 -6.39±0.09 2.32±1.81 -8.28±0.11

MPEA 1.37±0.10 -6.20±0.03 6.47±1.40 -8.03±0.03

MPF 11.56±0.19 -5.21±0.01 7.89±1.29 -6.70±0.02

Dexamethasone 1.69±0.03 -6.13±0.01 1.08±0.14 -7.94±0.01

DG 0.4±0.02 -6.77±0.03 1.22±0.26 -8.79±0.04

DA 1.38±0.04 -6.22±0.01 0.95±0.48 -8.05±0.02

DEG 0.72±0.03 -6.50±0.02 2.05±0.21 -8.44±0.02

DEA 0.97±0.02 -6.37±0.01 3.19±1.20 -8.25±0.01

DF 13.05±0.42 -5.15±0.01 7.15±1.18 -6.63±0.01

Betamethasone 1.94±0.18 -6.06±0.04 2.57±1.17 -7.85±0.05

BG 0.72±0.03 -6.51±0.02 0.90±0.51 -8.44±0.03

BEG 1.39±0.08 -6.20±0.02 4.24±0.81 -8.03±0.03

�

Tab

le 3

Stat

istic

al p

aram

eter

s cal

cula

ted

for A

NN

, PLS

and

MLR

mod

els.

aM

or16

s - si

gnal

16/

wei

ghte

d by

I-st

ate

(3D

MoR

SE d

escr

ipto

r);

nCt -

num

ber o

f to

tal t

ertia

ry C

(sp3

); C

-008

- C

HR

2X (A

tom

-cen

tred

fragm

ents

); C

ATS

2D_0

9_D

L - C

ATS

2D

Don

or-L

ipop

hilic

at l

ag 0

9.

bD

/Dtr0

6 - d

ista

nce/

deto

ur ri

ng in

dex

of o

rder

6;

CA

TS_2

D_0

8_LL

- C

ATS

2D L

ipop

hilic

-Lip

ophi

lic a

t lag

08;

F02

[C-C

] - fr

eque

ncy

of C

- C

at t

opol

ogic

al d

ista

nce

2; F

05[C

-

O] -

freq

uenc

y of

C-O

at t

opol

ogic

al d

ista

nce

5; M

LOG

P2- s

quar

ed M

orig

uchi

oct

anol

-wat

er p

artit

ion

coef

ficie

nt.

cnB

M -

num

ber o

f mul

tiple

bon

ds;

RD

F015

m -

Rad

ial D

istri

butio

n Fu

nctio

n- 0

15/w

eigh

ted

by m

ass;

RD

F145

s - R

adia

l Dis

tribu

tion

Func

tion-

145/

wei

ghte

d by

I-st

ate;

Mor

24s -

sign

al 2

4/w

eigh

ted

by I-

stat

e.

dH

OM

O -

high

est o

ccup

ied

mol

ecul

ar o

rbita

l; H

OM

T - H

OM

A to

tal (

geom

etric

al d

escr

ipto

r);

RD

F130

v - R

adia

l Dis

tribu

tion

Func

tion-

130/

wei

ghte

d by

van

der

Waa

ls v

olum

e;

CA

TS_2

D_0

5_A

L - C

ATS

2D A

ccep

tor-

Lipo

phili

c at

lag

05.

Mod

elR

egre

ssio

n eq

uatio

ns/s

elec

ted

desc

ripto

rsR

MSE

ER

MSE

PQ

2r

R2 pr

edR

2F

p

AN

N(lo

gPe)

logPe=

f (M

or16

s,nC

t,C-0

08,C

ATS

2D_0

9_D

L)a

0.10

20,

195

0.67

10.

922

0.86

7

PLS(

logPe)

logPe=

f(D

/Dtr0

6,C

ATS

2D_0

8_LL

,F02

[C-C

],F05

[C-O

],

MLO

GP2

)b

0.07

90.

172

0.96

10.

939

0.89

60.

974

34.8

150.

0000

1

MLR

(logPe)

logPe=

7.88

465+

0.00

009. M

or16

s+

0.52

423. C

-008

+0.

3901

2. CA

TS2D

_09_

DL

0.12

50.

264

0.89

50.

845

0.75

60.

917

56.4

37<0

.000

0

AN

N(R

)R=

f(nB

M,R

DF0

15m

,RD

F145

s,Mor

24s)

c0.

746

0.83

00.

888

0.95

00.

882

PLS(R)

R=f(

HO

MO

,HO

MT,

RD

F130

v,C

ATS

2D_0

5_A

L)d

1.10

91.

473

0.85

00.

853

0.63

00.

872

36.7

030.

0000

04

MLR

(R)

R=0.

9856

02. nB

M-0

,001

364. R

DF0

15m

+0,0

0006

9.

RD

F145

s+0,

0007

38. M

or24

s

0.51

90.

962

0.90

80.

955

0.84

20.

962

95.1

56<0

.000

0

��

�

Table 4

Sensitivity analysis of ANN(R) model.

Descriptor nBM RDF015m RDF145s Mor24s

Ratio a 3,19 1,60 1,29 1,78

a Sensitivity analysis parameter

��

�

��

�

Graphical abstract

17β-carboxamide steroids - In vitro prediction of human skin permeability and retention using PAMPA...

Documents

Transcript of 17β-carboxamide steroids - In vitro prediction of human skin permeability and retention using PAMPA...