Article No. jcht.1999.0541Available online at http://www.idealibrary.com on

J. Chem. Thermodynamics1999, 31, 1297–1306

Thermodynamic study of the imipramine–insulininteraction

Jose Luis Lopez-Fontan, Victor Mosquera, Carlos Rega, FelixSarmiento,Grupo de Fısica de Coloides y Polımeros, Departamento de Fısica Aplicaday Departamento de Fısica de la Materia Condensada, Facultad de Fısica,Universidad de Santiago de Compostela, E-15706 Santiago de Compostela,Spain

and Malcom N. Jonesa

School of Biological Sciences, University of Manchester, Manchester M139PT, U.K.

The interaction of the antidepressant drug imipramine with insulin in aqueous solution atpH= 3.2 (ionic strengthIc = 6.5·10−3 mol · dm−3) has been investigated by equilibriumdialysis and microcalorimetry. Imipramine binds to insulin to the extent of≈(50 to 60)imipramine molecules per insulin molecule in the temperature rangeT = (283.15 to308.15) K. The binding data have been used to obtain the Gibbs energy per imipraminemolecule bound1Gν using the Wyman binding potential approach. The enthalpy ofthe interaction is small and exothermic following a sigmoidal curve as a function ofbinding, and1Gν is dominated by a large entropy increase on binding which is of similarmagnitude to the non-specific hydrophobic binding of several other amphipathic ligandswith globular proteins. c© 1999 Academic Press

KEYWORDS: interaction; imipramine; insulin; thermodynamics

1. Introduction

The tricyclic antidepressant drug imipramine hydrochloride (IMI) has been shown to havea critical micelle concentration, and to form small micelles in aqueous solutions.(1–4) Italso forms mixed micelles with the related drug clomipramine,(5) and with the surfactantn-dodecyltrimethylammonium chloride.(6) In addition to its antidepressant activity ithas been shown to initiate other clinical effects, specifically affecting glucose levels inthe blood under insulin control,(7–10) and has been used in the treatment of diabetic

aTo whom correspondence should be addressed.

0021–9614/99/101297 + 10 $30.00/0 c© 1999 Academic Press

1298 J. L. Lopez-Fontanet al.

N

CH2CH2CH2NH(CH3)2 C1–+

neuropathy.(11) At moderate blood glucose levels(≈17 · 10−3 mol · dm−3) imipraminehas been found to suppress stimulated insulin secretion from mouse pancreatic isletsof Langerhans.(9) Imipramine also increases sensitivity to the hypoglycaemic effects ofinsulin.(7)

The ability of imipramine to form micelles suggests that it might interact with relativelyhydrophobic proteins such as insulin. In the present study, the interaction of imipraminewith insulin in aqueous solutions has been investigated. Equilibrium dialysis was used toobtain the extent of drug binding to insulin and hence the Gibbs energies of interaction.Microcalorimetry was used to obtain the enthalpies of interaction.

2. Experimental

MATERIALS

Crystalline insulin (molecular mass= 5734) from bovine pancreas (cat. no. 5500, 24International units per mg) and imipramine hydrochloride (C19H24N2HCl, molecular mass= 316.9) (cat. no. I 7379) were obtained from Sigma Chemical Co. and used as supplied.Dialysis membrane tubing (cat. no. D 7884, molecular mass cut-off= 2000) was fromSigma Chemical Co. All the other materials used were of the highest purity analyticalgrade available and solutions were made up in double-distilled, deionized, and degassedwater. The buffer was 50· 10−3 mol · dm−3 glycine hydrochloric acid, pH= 3.2, ionicstrengthIc = 6.5 · 10−3 mol · dm−3 at T = 298.15 K. The ionic strength of the bufferwas calculated from the glycine dissociation constant in acid solution (pKa = 2.3503 atT = 298.15 K).(12)

APPARATUS AND METHODS

Enthalpy measurements were made atT = 298.15 K using a Beckman 190 B twin-cell conduction calorimeter, described in detail previously.(13) Matched drop well glasscells were used; the sample cell was charged with 1· 10−6 m3 of insulin solution(mass concentration= 5 kg ·m−3) and 4· 10−6 m3 of imipramine solution of the requiredmolality (0 to 0.35) mol· kg−1 in a buffer. The reference cell was charged with 1·10−6 m3

of buffer and 4·10−6 m3 of imipramine solution of an identical concentration to that in thesample cell. When the solutions were mixed, the enthalpies of drug dilution cancelled, andthe enthalpy of insulin dilution was negligible.

The binding of the imipramine to insulin was measured by equilibrium dialysis inwhich 2 · 10−6 m3 aliquots of insulin solution of mass concentration 6.0 kg ·m−3

Insulin–imipramine interaction 1299

were placed in dialysis bags and equilibrated with 1· 10−5 m3 of imipramine solutionscovering the required concentration range. Dialysis was carried out over 192 h at thetemperatures (283.15, 288.15, 298.15, 303.15, and 308.15) K to attain equilibrium.Previous studies had established that this was a sufficient time for a low molar mass ligandto equilibrate across the dialysis membrane.(14) Over the temperature rangeT = (293.15to 308.15) K, the critical micelle molality of imipramine increases from 0.052 mol· kg−1

to 0.094 mol· kg−1.(3) The free imipramine concentrations at equilibrium were assayed bymeasuring the absorbanceA at 280 nm with a Beckman DU 640 spectrophotometer withreference to a standard curve which obeyed the following equation. The optical absorbanceA is defined as lg(I0/I )whereI0 is the incident light intensity andI is the transmitted lightintensity:

[IMI ]/(mol · dm−3) = (A− 0.023031)/20.09442. (1)

Each data point on the binding isotherms corresponds to an individual dialysis sample, andthe curves were made up of up to three batches of separate dialysis experiments to coverthe required range of free imipramine concentration.

The binding of the drug (IMI) to the insulin (In) can be represented by a sequence ofequilibrium as follows:

IMI + In⇔ In(IMI )1,

IMI + In(IMI )1⇔ In(IMI )2, (2)

IMI + In(IMI )2⇔ In(IMI )3.

In general, we may write for the formation of complexes in which, on average,ν drugmolecules are bound to the insulin:

IMI + ln(IMI )ν−1⇔ ln(IMI )ν . (3)

This sequence of binding reactions will have a corresponding sequence of equilibriumconstants, and associated Gibbs energies. For the formation of the average complexln(IMI )ν as represented by equation (3), the Gibbs energy per drug molecule bound isdefined as1Gν .

To obtain the Gibbs energies per drug molecule bound(1Gν), the isotherms were fittedwith polynomials of the form:

ν = a1+ a2 ln([IMI ]f/mol · dm−3)+ a3 ln([IMI ]f/mol · dm−3)2+

a4 ln([IMI ]f/mol · dm−3)4, (4)

whereν is the average number of imipramine molecules bound per monomeric insulinmolecule and[IMI ]f is the free imipramine concentration.

Third-order polynomials were chosen because they gave the highest correlation coef-ficients and the lowest standard errors for the whole data set. Outside the data range thepolynomials have no physical significance. Table1 summarizes the statistics of the fit. Thepolynomials were used to calculated the Wyman binding potentials(5) as a function ofνby using the following equations:(15)

5 = RT∫ ln([IMI ]f/mol·dm−3)ν=ν

ln([IMI ]f/mol·dm−3)ν=0

νd ln([IMI ]f/mol · dm−3), (5)

1300 J. L. Lopez-Fontanet al.

TABLE 1. Statistics of the polynomial fits of insulinbinding isotherms for imipramine at pH= 3.2 anddifferent temperaturesT, r is the correlation coefficient,σ is the standard error of the fit ofν, anda1, a2, a3, and

a4 are the coefficients of equation (4)

T/K r σ(ν) a1 a2 a3 a4

283 0.864 4.38 854 547 118 8.4

288 0.784 6.83 270 126 19 0.94

298 0.840 4.94 769 523 120 9.20

303 0.947 5.40 1511 991 220 16.3

308 0.933 4.59 833 496 99 6.50

–5.5

0

5

10

15

20

25

30

35

40

45

50

55

–5.0 –4.5 –4.0

ln ([IMI]free / mol·dm–3)

–3.5 –3.0 –2.5

v

FIGURE 1. Binding isotherms{number of imipramine molecules bound per protein molecule(ν)

as a function of ln([IMI ]free/mol · dm−3)} for the interaction of imipramine with insulin in aqueoussolution at pH= 3.2, Ic = 6.5 · 10−3mol · dm−3; •, T = 283.15 K,∗, T = 288.15 K. The insulinconcentration was 6.0 kg ·m−3. The c.m.c. of imipramine atT = 298.15 K is indicated by the arrow.

Insulin–imipramine interaction 1301

TABLE 2. Number of imipramine molecules per insulinmolecule(ν) as a function of free imipramine concentra-tion ([IMI] free) in equilibrium with the imipramine–insulincomplexes at the temperatures (283.15, 288.15, 298.15,

303.15, and 308.15) K

ln [IMI ]free ν ln [IMI ]free ν ln [IMI ]free ν

T = 283.15 K

−3.10 35 −3.86 5 −4.35 8

−3.17 31 −3.99 9 −4.50 5

−3.29 32 −4.12 11 −4.88 6

−3.47 25 −4.14 9 −5.04 7

−3.64 11 −4.18 5 −5.15 7

−3.79 9 −4.32 10

T = 288.15 K

−2.93 51 −3.55 33 −4.44 6

−2.95 39 −3.59 22 −4.79 10

−3.02 36 −3.67 14 −4.96 8

−3.10 35 −3.92 19 −5.24 8

−3.23 39 −4.40 14

−3.41 33 −4.50 15

T = 298.15 K

−2.85 48 −3.64 11 −4.69 13

−2.90 30 −3.81 11 −4.83 11

−2.99 31 −3.91 15 −5.05 14

−3.13 39 −3.97 14 −5.18 10

−3.24 26 −4.12 18 −5.26 9

−3.40 16 −4.19 16

−3.54 16 −4.34 11

T = 303.15 K

−3.04 72 −3.97 24 −4.58 18

−3.38 61 −4.14 27 −4.72 20

−3.41 49 −4.16 26 −4.74 18

−3.43 35 −4.21 23 −4.85 18

−3.54 33 −4.35 24 −4.95 14

−3.61 24 −4.38 22 −4.99 12

−3.73 19 −4.49 21 −5.32 13

T = 308.15 K

−3.08 48 −3.93 11 −4.69 9

−3.18 48 −4.03 12 −4.85 12

−3.29 49 −4.24 15 −5.01 13

−3.39 46 −4.39 13 −5.30 13

−3.39 30 −4.43 12 −5.30 13

−3.49 27 −4.44 9 −5.36 11

−3.64 27 −4.65 14

1302 J. L. Lopez-Fontanet al.

–5.5

0

10

20

30

40

50

60

70

80

–5.0 –4.5 –4.0

ln ([IMI]free / mol·dm–3)

–3.5 –3.0 –2.5

v

FIGURE 2. Binding isotherms{number of imipramine molecules bound per protein molecule(ν) asa function of ln([IMI ]free/mol · dm−3)} for the interaction of imipramine with insulin in aqueoussolution at pH= 3.2, Ic = 6.5 · 10−3mol · dm−3; N, T = 298.15 K; �, T = 303.15 K;�, T = 308.15 K. The insulin concentration was 6.0 kg ·m−3. The c.m.c. of imipramine atT = 298.15 K is indicated by the arrow.

5 = RT

{a1 ln([IMI ]f/mol · dm−3)+ (a2/2)(ln [IMI ]f/mol · dm−3)2+

(a3/3) ln([IMI ]f/mol · dm−3)3+ (a4/4) ln([IMI ]f/mol · dm−3)4+ · · ·

}.

(6)

whereR is the universal gas constant. The equilibrium constantsK for the sequence ofνbinding reactions described above were calculated by using the equation:

5 = RT ln(1+ K [IMI ]νf /mol · dm−3). (7)

And hence

1Gν = −(RT/ν) ln K , (8)

where K is a function ofν, and the Gibbs energy(1Gν) is an average value for thesequence of binding reactions leading to a complex with an average composition In(IMI )ν .It has not been defined as a standard Gibbs energy. The c.m.c. of imipramine in thepH= 3.2 buffer was determined from conductivity measurements made with an HP 4285APrecision LCR Meter equipped with an HP E5050A Colloidal Dielectric Probe operatingat a frequency of 200 kHz. To obtain the highest precision, the probe was housed in aspecifically designed cell so that the probe head was always filled with at least 0.02 m

Insulin–imipramine interaction 1303

TABLE 3. The massic enthalpy1hν of in-sulin interaction as a function of the molalityof imipramine, data plotted in figure3 at T =

298.15 K

m(IMI )/(mol · kg−1) 1hν/(kJ · g−1)

0.010 −0.3

0.020 −0.5

0.025 −0.7

0.030 −1.0

0.035 −2.1

0.040 −2.0

0.045 −4.5

0.045 −5.1

0.048 −4.7

0.060 −7.7

0.070 −8.2

TABLE 4. The Gibbs energy1Gν , enthalpy1Hν , and entropy expressed asT1Sν , forthe specific (s) and non-specific (ns) interactions of amphipathic ligands with proteins at

T = 298.15 K

Protein liganda pH ν1Gν

(kJ ·mol−1)

1Hν(kJ ·mol−1)

T1Sν(kJ ·mol−1)

Ribonuclease-OBG (ns) 4 100 −11.2 0.311 11.5

Ribonuclease-SDS (s) 3.2 18 −29.6 1.26 30.9

Lysozyme-OBG (ns) 6.4 130 −10.2 0.783 11.0

Lysozyme-SDS (s) 3.2 16 −26.8 8.4 18.5

Glucose oxidase-SDS (ns) 3.2 575 −17.0 −3.34 13.7

Glucose oxidase-SDS (s) 3.2 120 −21.2 −3.29 17.9

Glucose oxidase-DTAB (ns) 10.0 510 −15.8 0.845 16.7

Catalase-SDS (s) 6.4 331 −28.4 −8.36 20.0

Catalase-SDS (ns) 6.4 500 −21.5 −3.5 18.0

Insulin-SDS 3.2 20 −17.0 −3.07 13.9

Insulin-imipramineb 3.2 25 −11.3 −1.12 10.2

a OBG (n-octyl-β-D-glucopyranoside), SDS (sodiumn-dodecylsulphate) DTAB (n-dodecyl-trimethylammonium bromide),b This work, the rest of the data was taken from reference 16.

depth of solution. The cell with the probe was immersed in a Techne, model RB-12Athermostatted bath equipped with a Tempunit TV-16A. Temperature control was achievedusing an Anton Paar DT 100-30 thermostat to maintain the temperature constant to within

1304 J. L. Lopez-Fontanet al.

0.00–10

–8

–6

–4

–2

0

2

0.01 0.02 0.03 0.04

m(IMI)/ (mol·kg–1)

0.05 0.06 0.07 0.08

1h

/ (kJ

·g–1)

0 5 10 15 20 25 30 35v

FIGURE 3. The massic enthalpy of interaction(1h/mol · kg−1) of imipramine with insulin aqueoussolution as a function of the final imipramine molality and of the number of molecules of imipraminebound per insulin molecule(ν) at pH= 3.2, Ic = 6.5 · 10−3mol · dm−3; T = 298.15 K. The finaltotal insulin concentration was 1.0 kg ·m−3.

±0.01 K. The solution was agitated using a shaker. The c.m.c. was obtained from theintersection of the lines of conductivity against concentration. The resulting break point(c.m.c.) was found to be 0.51 mol· kg−1 at T = 298.15 K.

3. Results and discussion

The observed values of the number of bound imipramine molecules per insulin molecule atseveral temperatures are given in table2. The binding isotherms for imipramine binding toinsulin at pH= 3.2 and the temperatures (283.15, and 288.15) K are shown in figure1, andfor the temperatures (298.15, 303.15, and 308.15) K in figure2. The isotherms relate to freeimipramine concentrations below the c.m.c.s. The characteristic features of the isothermsreside in steep rises inν as the c.m.c.s are approached. The increases are more evident withincreases in temperature, indicating cooperativity of binding at higher free imipramineconcentrations. The number of imipramine molecules bound per protein molecule canreach 50 to 60, although the variation of binding with temperature is too small to enable

Insulin–imipramine interaction 1305

10–15

–10

–5

0

5

10

15

15 20 25 30 35

v

(1G

v, o

r 1

Hv,

or

T1

S v) /

(kJ·m

ol–1

)

FIGURE 4. The Gibbs energy1Gν , enthalpy1Hν , and entropy1Sν expressed asT1Sν , for theinteraction of imipramine with insulin as a function of the number of molecules of imipramine boundper insulin molecule(ν) at pH= 3.2, Ic = 6.5 · 10−3mol · dm−3, andT = 298.15 K.�,1Gν ; N,1Hν ; •, T1Sν .

the enthalpy of binding to be determined accurately from the temperature coefficient ofbinding by the van’t Hoff methods.

The enthalpy of interaction of imipramine with insulin at pH= 3.2 andT = 298.15 Kwas measured directly by microcalorimetry. The results are shown in table3, and plotted infigure3. The upper scale on this figure shows the equilibrium binding levels correspondingto the total imipramine concentration (lower scale). The enthalpy data is sigmoidal withrespect to imipramine concentration. The values ofν were calculated from a plot ofνagainst [IMI]total, for which [IMI] total was calculated by using the equation:

[IMI ]total = [IMI ]free+ ν[Insulin], (9)

with reference to the binding data, in which the insulin concentration was equal to thefinal concentration of insulin in the calorimetric experiments. From the data in table3,the enthalpy per imipramine molecule bound(1Hν) was determined and combined with1Gν to obtain the entropy change per imipramine molecule bound(1Sν). Figure4 shows1Gν , 1Hν , andT1Sν as a function ofν. It is clear that the dominant contribution to1Gν is a large and positiveT1Sν term consistent with hydrophobic interaction. It wasnot possible to extend the measurements to other pH conditions since it was found thatabove pH= 4 insulin was not soluble until higher pHs (8 to 9) at which imipramine was

1306 J. L. Lopez-Fontanet al.

insoluble. We were thus unable to obtain conditions where both components were solubleeven after manipulating the solution ionic strength and using alternative buffer systemssuch as phosphate and borate.

Table 4 shows the thermodynamic parameters for the interaction of amphipathic lig-ands with several proteins taken from previous studies.(16) The table shows the thermo-dynamic parameters for both the non-specific (hydrophobic) and specific (ionic) interac-tions of ligands with proteins. In general, the Gibbs energies of the specific interactionsare more negative than those of the non-specific interactions, largely as a consequenceof the ionic interactions between positively charged proteins and negatively charged lig-ands, such as sodiumn-dodecylsulphate (SDS), or in the case of glucose oxidase andn-dodecyltrimethylammonium bromide (DTAB) at pH= 10, the negatively charged pro-tein and the cationic ligand. At pH= 3.2, both insulin and imipramine are positivelycharged and will interact mainly hydrophobically, hence the thermodynamic parametersfor the insulin–imipramine interaction are similar in magnitude to those found for the non-specific interactions between ligands and proteins. The enthalpies of interaction play a rela-tively minor role in determining the values of1Gν which are dominated by large increasesin the entropy on binding that are characteristic of a significant hydrophobic contributionto the interactions.

REFERENCES

1. Moro, M. E.; Rodriguez, L. J.Langmuir1991, 7, 2017–2020.2. Attwood, D.; Boitard, E.; Dubes, J.-P.; Tachoire, H.J. Phys. Chem.1992, 96, 11018–11022.3. Attwood, D.; Blundell, R.; Mosquera, V.; Garcıa, M.; Rodrıguez, J.Colloid Polym. Sci.1994,

272, 108–111.4. Attwood, D.; Mosquera, V.; Garcia, M.; Suarez, M. J.; Sarmiento, F.J. Colloid Interface Sci.

1995, 175, 201–206.5. Sarmiento, F.; Lopez-Fontan, J. L.; Prieto, G.; Attwood, D.; Mosquera, V.Colloid Polym. Sci.

1997, 275, 1144–1147.6. Casarotto, M. G.; Craik, D. J.J. Phys. Chem.1992, 96, 3146–3149.7. Kathol, R. G.; Jaeckle, R.; Wysham, C.; Sherman, B. M.Psychiatry Res.1992, 41, 45–48.8. Gupta, B.; Awasthi, A.; Jaju, B. P.Ind. J. Med. Res. B1992, 96, 65–71.9. Eldakhakhny, M.; Ellatif, H. A. A.; Ammon, H. P.T. Arzneimittel-Forschung1996, 46, 667–669.

10. Pilj, H.; Meinders, A. E.Drug Safety1996, 14, 329–333.11. Sindrupo, S. H.Danish Med. Bull.1994, 41, 66–78.12. Handbook of Chemistry and Physics:57th edition. de West, R. C.: editor. CRC Press: Boca

Raton, FL.1976, p. D147.13. Pilcher, G.; Jones, M. N.; Espada, L.; Skinner, H. A.J. Chem. Thermodynamics1969, 1, 381–

392.14. Prieto, G.; del Rio, J. M.; Paz Andrade, M. I.; Sarmiento, F.; Jones, M. N.Int. J. Biol. Macromol.

1993, 343–345.15. Wyman, J. J.J. Mol. Biol.1965, 11, 631–640.16. Jones, M. N.Surface Activity of Proteins: chemical and physiochemical modifications. Mag-

dassi, S.: editor. Marcel Dekker, Inc.:1996,Chapter 8, p. 237.

(Received 13 August 1999; in final form 26 April 1999)

WA98/043