HYBRIDOMA Volume 11, Number 3, 1992 Mary Ann Liebert, Inc., Publishers

Kinetic and

Energetic Parameters of Imipramine Binding to Monoclonal Antibodies as Measured by Fluorescence Spectroscopy DOUGLAS GABLER, CHHABINATH MANDAL, CHARLES HARRINGTON, MACIEJ ADAMCZYK, and D. SCOTT LINTHICUM Center for Macromolecular Design and the Department of Veterinary Pathobiology, College of Veterinary Medicine, Texas A&M University, College Station, TX 77843-4467 Abbott Laboratories, Chemical Neurodiagnostics, Abbott Diagnostics Division, Abbott Park, IL 60064

ABSTRACT Monoclonal antibodies which bind small drugs are useful for the study of the interactive forces involved in antibody-ligand complexation. Detailed understanding of these supramolecular forces requires a careful examination of structural and Fluorescence spectroscopy thermodynamic parameters of the interacting molecules. techniques are very useful in this regard. We report here, the kinetic and energetic parameters of four monoclonal antibodies made against the tricyclic antidepressant imipramine. These monoclonal antibodies were found to possess high to very high binding affinity constants, ranging from 107 to 10*eM"*, and caused fluorescence quenching or enhancement of a fluorescein labelled imipramine. The dissociation rates of the fluorescent ligand from the complexes were measured at different temperatures in order to provide some insight regarding the kinetic and energetic (thermodynamic) parameters of the antibody-ligand binding interactions.

INTRODUCTION antibodies (Mab) which bind small ligands and demonstrate do so because of the precise chemical interactions with individual amino acid side chain moieties which make up the antibody binding pocket An understanding of the interactive forces involved in antibody-ligand (1). complexation is important to our knowledge regarding the specificity of antibodies and immune recognition phenomena. Structural, thermodynamic, and kinetic experimental results, interpreted with quantum mechanical and statistical mechanical theory, are essential to this goal. The mere measurement of a binding affinity constant is, in most cases, not sufficiently robust to identify the binding forces responsible for the antibody-ligand complexation. It is, of course, an oversimplification to claim that a few empirical observations will identify all the interactive forces involved, but it may be reasonable to draw some inferences that are consistent with theoretical concepts. X-ray diffraction studies of protein crystals derived from myeloma proteins and Mab Fab fragments have demonstrated that the antibody binding site, in almost all cases, is created by six peptide loops at the tip of the Fv immunoglobulin fragment (2,3). These six loops, also known as "complementarity determining regions" (CDR), are generated by the folds and turns of the anti-parallel beta-stranded barrels or "sandwiches" of the heavy and light chain variable regions (4). Specific side chain moieties for the amino acids of the CDR provide the basis for the electrostatic, hydrogen bonding, van der Waals, - charge complex, and hydrophobic interactions with the antigen. A sub-set of "energetic" antibody epitopes in the CDR may account for the majority of the attractive Monoclonal

stereospecific binding

301

forces involved in antigen binding, and these epitopes are, of course, complemented by favorable interactive epitopes in the antigen (5). The quantitative measurement of the reaction kinetics, such as association (k+i) and dissociation (k.i) rate constants, and thermodynamic driving forces involved in the complexation, may be more useful than less precise terms, such as "affinity", because these aspects provide insight into the chemical nature of the ligand-receptor interactions. Studies of ligand complexation with polyclonal antibodies or Mab (6-8) have shown that both enthalpic and entropie factors make significant contributions to the In the present study, we have used fluorescent spectroscopy binding energetics. techniques to examine the kinetics and energetics of ligand binding for several Mab which bind the neurogenic Fluorescence antidepressant drug imipramine. spectroscopy has several advantages for these types of experiments: i) a high degree of sensitivity, ii) a covalently linked fluorescent molecule which acts as a "reporter" when the ligand binds to the antibody, and iii) intensity of the fluorophore can be attenuated (quenched) by the local molecular environment of the larger antibody molecule. MATERIALS AND METHODS

Imipramine

and

Fluorescent-Imipramine

The fluorescent ligand, imipraminefluorescein #37241-36-C (IMIF), was produced Abbott Laboratories as a stock solution (6.27xlO"5M) in 0.1 M glyclglycine buffer. IMIF was diluted to 3.14 xl0"5M with phosphate buffered saline, pH 7.4, (PBS), and stored The fluorescein marker was covalently linked to imipramine at 4°C in dark bottles. through the C2 carbon of the tricyclic ring. Imipramine (IMI) was also obtained from Abbott Laboratories as a solid and a stock solution (6.12xlO"^M) was made in PBS and stored as above prior to use. at

Preparation

of Monoclonal

Antibody

(Mab)

laboratory using our standard For the study presented herein, we 5.3, 5.4, and 5.5, which were all made against imipramine conjugated to (bovine serum albumin) through the C2 ring position (same as the secreting hybridoma clones were identified by a radioimmunoassay using ligand; the alkyl chain of IMI has been identified as the immunodominant molecule for the antibodies used in this study. Further studies to identify with other tricyclic are underway.

MAb to imipramine were produced in hybridoma procedures described previously (9).

used Mab 5.1, carrier protein IMIF). Mab 3H-IMI as the portion of the

cross-reactivity

our

Fluorescence_Measurements All fluorescence measurements were made using a SLM 8000C spectrofluorometer. For these experiments the fluorescence was monitored at an excitation wavelength of 490 nm (8 nm bandpass) and the emission monochrometer set at 517 nm (4 nm bandpass) at a 90° angle relative to the excitation light. In order to compensate for any fluctuation in the lamp intensity, the relative fluorescence intensities were obtained as the ratio of a fluorescent beam to a straight-through beam. The temperature was maintained with a thermostable sample compartment at T+0.5° C. At the wavelengths used no correction was needed for protein fluorescence because proteins neither absorb nor emit at these

wavelengths.

Determination of Maximum Fluorescence

Signal

Spectra were acquired at a constant excitation and emission wavelength for 20 seconds. The area under the curve was then determined to provide relative integrated intensities for each spectrum. The maximum fluorescence or signal, Qm a representing 100% binding of the fluorescent marker was determined by titrating antibody into a solution of fluorescein derivatized antigen (IMIF). Solutions of IMIF (3.0 ml; 1.047 10_8M) were titrated with Mab (10_5M) in 1-5 µ steps until typically 10-30 µ had been added. The resulting titration graph exhibited a smooth curve that very nearly leveled out. A typical titration plot is shown in Figure 1. The total volume of >

302

10

20

40

30

50

Antibody (µ . added) Figure verses

1 Titration of IMIF with Mab 5.3 at 35°C plotted relative fluorescence intensity (arbitrary units).

as

volume of

antibody added

was less than 1% of the volume of the solution being titrated and control Such very small additions did not corrections were not necessary. showed experiments disturb the equilibrium of the system. Antibody TM-1 (IgGl, kappa) was used as a negative control and demonstrated no non-specific binding of IMIF. Qmax was calculated as follows: Qmax = U-(Fo/Foo) I, where F0 is the integrated intensity of the spectrum before the addition of Mab and F^o is tne intensity at 100% binding of IMIF. The intensity at 100% binding (F^) was determined from a double inverse plot of 1/vol verses 1/intensity for the last 3-5 data points from the nearly level region of the titration plot. A typical plot is shown in Figure 2. In the double inverse plot l/(y-intercept) represents Foo or 100% binding.

titrant added

Determination of

Using the

Binding Affinity

Constants

fluorescence spectroscopy data Scatchard

analysis (10)

was

performed

0.179 c -

£

0.178

0.177

o 3

0.176 0.02

1/Volume

Double inverse

Figure

2

35°C

(Fig. 1)

intensity

at

as

100%

plot

1/volume

binding

0.04

0.03

of data

versus

(µ )

points obtained

in titration of IMIF with Mab 5.3 at

1/fluorescence intensity; 1/y intercept =l/fluorescence

of IMIF and the

y-intercept 303

=

0.183

CD

4

5 mAb

6

7

10-8 /

Stinson-Holbrook analysis of Mab 5.3 binding with IMIF at 35°C. The relative fluorescence intensities were obtained from the titration of IMIF with Mab 5.3 (Fig. 1). The slope = Ka = 5.15 xlO8

Figure 3

each of the Mab-IMIF complexes, but these experiments proved to be very time consuming. We determined that the method developed by Stinson and Holbrook (20) was less susceptible to experimental error (reproducibility of data points), provided more data points and was far less time consuming. A comparison of the results obtained by S cate hard and Stinson analysis proved that the two methods provided similar (within 10A another 20%) affinity determinations for several of the Mab used in this study. advantage of Stinson-Holbrook analysis was that corrections for inner filter effects were not necessary because the proteins (the titrant in these experiments) did not absorb at the wavelengths employed. We have compared the results obtained with a radioimmunoassay using 3H-IMI and found the affinity constant determinations to be in agreement with spectroscopic data. Stinson-Holbrook analysis was performed on the data obtained during the Qmax determinations described above. A plot of [titrant]/q vs.l/(l-q) yields as the slope Ka; where q AF0/AFmax. AF0 Fn Fi and AFmax Ff F¡; where Fn Fj and Ff are the nth, initial and final intensity values, respectively. A typical Stinson plot is shown in Fig.3.

on

=

=

=

-

,

,

Dissociation Rate Determinations

Samples were prepared by adding 1.0 µ IMIF (3.14xlO"^M) to 3.0 ml PBS and a spectrum acquired. Antibody (3-15 µ ; 10"6 M) was then added and a spectrum acquired after 30 min. Unlabeled antigen (IMI, 1.5-3.0 µ ; 6.12xlO"3M) was then added and the The spectra were acquired for 40-500 seconds at a constant spectra was measured. excitation and emission wavelength. The area under the curve was then integrated in 2 second steps for the length of the experiment to obtain relative integrated intensities which were used in a computer program to determine rate constants. The binding of the Mab to IMIF caused an intensity change of about 40% of the ligand fluorescence for all the Mab studied. Mab 5.1, 5.4, and 5.5 caused an enhancement of the IMIF fluorescence, while Mab 5.3 caused a quench of the fluorescent intensity of IMIF. The subsequent addition of unlabeled imipramine resulted in a time dependent reversal of the fluorescence intensity. It should be noted that these Mab have nearly the same affinity constants for IMI and IMIF (preliminary determinations using a radioimmunoassay at room temperature) and this is due, in part, to the fact that the fluorescein label is located at C2 and is in the same position as the chemical conjugation of the eliciting immunogen. With these apsects in mind, the change in the IMIF intensity follows a pseudo-first order kinetics:

304

< c

1

2

Time (min)

Figure.

4

Dissociation of IMIF from Mab 5.3 at 25°C.

A=fluorescence

time, x=initial fluorescence intensity before addition of IMI.

Antibody (Ab)

+

intensity

at

nth

k+1

Ligand (L)

Complex (Ab-L)

->

k-1 For most antibody ligand systems, the association rate constant (k+i) is generally within one decimal order of magnitude of 108 lM^sec"1 because this rate approaches the theoretical limiting value of diffusion-controlled reactions. The dissociation rate constant (k_i), reflects the stability of the complex. A typical kinetic plot (Mab 5.3) is presented in Figure 4. The dissociation rate constants and half-lives were calculated using an algorithm developed in our laboratory for the Macintosh IIx which utilizes the standard relaxation equation (12). Binding

Calculations

Standard free energy (AG°) of binding -RTlnKa ; where R is the gas constant,

calculated using the equation: is the absolute temperature and Ka is the 0 (standard enthalpy) was obtained from a van't Hoff plot of 1/T affinity constant. versus lnKa; where the slope is equal to -AH°/R. AS0 (standard entropy) is obtained from the equation: AG° = 0 TAS0.

AG°

was

=

-

Temperature_Dependence The temperature dependence of the dissociation rates were studied in the range between 15°C and 40°C. The enthalpy ( #), free energy (AG#) and entropy (AS#) of activation were calculated according to the following equations (13):

d(lnk

,)

d(L)

AG /RT

. -1

R

h

and =AE-RT

áG

305

a

=

a -

TAS

it

where is the absolute temperature, R is the gas constant, k is the Boltzman's constant, h is Plank's constant and was obtained from is the measured energy of activation. the

slope of

the

plot

of

ln(k i); where the slope is equal

vs —

to

- /R.

RESULTS AND DISCUSSION The

dissociation rate constants, half-lives, binding constants, free energies, and entropies of the imipramine-antibody complexes are presented in Table 1. The four monoclonal antibodies in this study bind imipramine with different affinities and dissociation rates. The dissociation kinetics of the ligand were studied at various temperatures and the initial dissociation rates for all the antibodies followed first order kinetics.

enthalpies,

TABLE 1

Kinetic and Mab

Temp (°K)

k.

sec"1

Thermodynamic t1/2 sec

Parameters of the MAb-IMIF

AG°

Ka M"1

0

complexes TAS0

kcalM^deg"1 kcalM^deg"1 kcalM"1

5 288 298

1.61e-4 5.74e-3

4300 1210

_203_l,83e-3_379

3.27e+7 6.19e+7 7.79e+7

-9.90 -10.62

4.62e+8 5.79e+8 l.He+9

-12.01 -12.35 -12.95

+7.68 +7.68

17.63 18.24

-11.12_£L6J_IMS_

5.3 303 308 313

1.63e-3 2.71e-2

298 303 308

6.33e-4 1.13e-3

298 303 308

7.57e-4 1.34e-3

42.5 25.6

7.43e-2_9.3

+16.41 +16.41

28.36 28.83

+16.41_29.30_

5.4 1090 613

2.13e-3_325

9.14e+8

-12.22 -12.39

8.62e+8 7.29e+8

-4.10 -4.10

8.11 8.24

-12.49_I4J0_838_

5.5

915 517

2.01e-3_345

1.69e+9 1.01e+10 1.52e+10

-12.58 -13.87 -14.35

+40.22 +40.22

52.92 53.81

+40.22_54.70_

As shown in Table 1 all of the Mab-complexes have negative AG° values and are in with their affinity constants. Most antibody-ligand reactions are 0 is generally negative (14). exothermic, in that Exceptions that are endothermic (positive 0) have been observed, such as that for anti-Rh0 (D) antibodies and Rh0 (D) positive erythrocytes (15). In endothermic reactions, AS0 must be large and positive in order to counteract the positive 0 or else the complexation will not take place. As presented in Table 1, Mab 5.4 is the only one of the four Mab studied that shows a 0 (exothermic) upon ligand binding; it is also the only Mab that shows a negative decreasing Ka with increasing temperature. Mab 5.1, 5.3 and 5.5 have endothermic reactions. The TAS0 of Mab 5.1, 5.3, and 5.5 are large and positive, and therefore compensate for the positive 0 of these systems. The positive ° of 5.1, 5.3, and 5.5 suggests a conformational change of the antibody takes place upon binding the ligand (16). The exclusion of ordered solvent (water) from side chains in the binding pocket of the antibody may be responsible for a portion of the large entropy component (TAS0). The strong tendency of apolar groups to self-associate through van der Waals forces is hampered by the presence of water and it has been reported that this prevents interactive moieties from approaching more closely than about 20 Â (17). The movement of ordered water out of the binding site provides for a better fit between the interactive moieties. Mainchain and side chain rearrangements in the MAb may also contribute in terms of a conformational entropy.

agreement

306

Mab 5.4 has a negative 0 and a TAS0 that is substantially lower than the other three Mab complexes studied. This leads us to believe that this reaction is enthalpically driven by new bond formation such as -bonds or electrostatics. Because imipramine has a protonated amine on the alkyl chain (pKa = 9.5) it is likely that this portion of the ligand is involved in an electrostatic or "salt-bridge" interaction with a negatively charged glutamate or aspartate residue in the antibody binding pocket. Such types of interactions have been identified in other antibodies which bind alkaloid drugs (18). Although there are often hydrogen bonds between the appropriate antibody side chains and water, the net enthalpy change due to an electrostatic bond makes a significant contribution to the complex. Dissociation rates of IMIF from the antibodies were measured at a number of different temperatures. Arrhenius plots (1/T versus Ink. ) are straight lines as shown in Figure 5. The thermodynamic activation (#) parameters (e.g., free energy, enthalpy and entropy of activation) were calculated using the equations presented above and these values are presented in Table 2. The free energies of activation of the complexes with 5.1, 5.4, and 5.5 are almost the same, but their enthalpies of activation differ due to differences in the entropie contributions. The complex with 5.3 showed the largest enthalpic barrier, but a positive entropie contribution, thereby yielding a free energy of activation lower than those of the other three antibody-ligand complexes. The dissociation rate for the complex with 5.3 was determined to be about 40 times faster than those for the other Mab. The energy differences corresponding to the state of activation, i.e., A+B to AB, presented in Table 2, correspond to the process of activation in the release of the ligand from the binding pocket of the antibody. These values account for the activation barrier in going from the stable complex AB to the high energy activated state AB* before breaking apart from the complex. These parameters are useful to describe the chemical environment in the immediate vicinity of the ligand binding pocket. Any change in the binding site would be reflected in the energy terms resulting in the acceleration or deceleration of the reaction velocity and, subsequently, its temperature dependence. For the Mab 5.1, 5.4, and 5.5 the free energies of activation ( #) are essentially the same, but their enthalpic and entropie contributions appear to be different. The enthalpy term arises from the bonding contributions (e.g., stacking, van der Waals, hydrogen bonding and other electrostatic energies) while the entropy term arises from the alteration in the extent of solvation, and the change in the rotational and vibrational freedom of the side chains of the residues. The higher enthalpy of activation for the complex with 5.3 is balanced by the positive entropie contribution making # slightly lower than for the other three complexes. The differences in the entropie terms would most likely arise from both aromatic side chain

c

3.1

3.2

3.3

1/Temp (°K

Fig. In k.

5

3.4

X10-3)

Arrhenius plot of dissociation of IMIF from Mab 5.4 at various temperatures; is the natural log of dissociation rate constant; slope= - /R = -1.03 10"*

307

TABLE2

Thermodynamic Parameters of AG* Temp kcal mole"1 deg" * (°K)

Mab

the Activated

AH* kcal

mole"1

Complexes

deg"1

TAS* mole"1

kcal

5.1 278 288 298

21.34 21.84 21.86

16.49 16.47 16.45

-4.85 -5.37 -5.41

308_21.91_16.43_-5.48 5.3

298 303 308

20.31 20.22 20.26

25.49 25.48 25.47

21.80 21.83 21.81

19.14 19.13 19.12

5.17 5.25 5.21

-ÌL2_1&21_2SAL·_149_ 5.4

298 303 308

-2.66 -2.70 -2.69

jm_1L32_L2JJ_-2,81

5.5

298 303 308

21.70 21.73 21.85

17.23 17.22 17.21

-3.54 -3.93 -4.32

_212_22.35_17.20_4J\_ and solvation, because the binding sites of the antibodies are thought to The energy of activation for the complex with 5.3 is water molecules. lower, thereby creating an off rale which is about 40 times faster than that observed for the other three antibodies. In this case, the process of activation is accompanied by a positive entropie effect (Table 2), implying decreased solvation and/or increased rotational or vibrational motions in the activated state. Biophysical measurements of the ligand-antibody interactions and competitive binding studies can prove useful in understanding the chemical make-up and intermolecular forces that participate in the ligand-antibody interaction. The measurement of an association constant or "affinity" for an individual antibody-ligand system as derived from a binding assay does not provide adequate information regarding these aspects. Kinetic and thermodynamic analyses can often provide clues regarding the enthalpic or entropie forces which are involved in the stabilization of the ligand-receptor complex (14). Analysis of the thermodynamic aspects of ligandantibody interactions have proven useful in understanding the driving forces and the molecular features for other small ligands, such as 2,4-dinitrophenyl; the binding energy to polyclonal antibodies or the myeloma protein 315 was determined to be primarily contributed by enthalpic factors (6). Subsequent analysis of the amino acid sequences from these antibodies revealed that several tryptophans in the binding site could be in contact with the hapten, and perhaps as much as -10 kcal/mol of enthalpy change could be ascribed to - charge transfer. Magnetic resonance techniques used to refine a model of the antibody binding site revealed an "aromatic box" with stacking interactions between the dinitrophenyl ligand and L93:W of the antibody (7). The molecular interactions which take place between a ligand and the antibody binding site are predicated on the precise location and identity of the amino acids that make up the antibody CDR loops. The size, shape and orientation of the binding site residues provide the stereospecificity and energetics of ligand binding. In addition, the ligand possess a set of complementary interactive epitopes. These aspects may prove very important in the design and engineering of antibodies with defined specificities and performance characteristics. In addition, a better understanding of the thermodynamic parameters involved in antibody-ligand complexation may aid in the the design of new drugs and analytes (19). movements

contain

numerous

308

ACKNOWLEDGMENTS We thank Ms. Francine D. Shirley for technical assistance in the preparation of the monoclonal antibodies. This work was supported by grants from the American Heart Association (891066) and the NIH/NIDA (DA07240).

REFERENCES 1. Novotny, J., Bruccoleri, R., Newell, J., Murphy, D., Haber, E. and Karplus, M. (1983) Molecular anatomy of the antibody binding site. J. Biol. Chem. 258:14433. 2. Davies, D. R., Sheriff, S. and Padlan, E. A. Chem. 263:10541. 3. Davies, D. R. and Immunol. 1:87.

Metzger,

H.

(1988) Antibody-antigen complexes. J. Biol.

(1983) Structural basis of antibody function. Ann. Rev.

Kabat, . ., Wu, T. T. and Bilsofsky, H. (1977) Unusual distributions of amino acids in complementarity determining (hypervariable) segments of heavy and light chains of immunoglobulins and their possible roles in specificity of antibody-combining sites. J.

4.

Biol. Chem. 252:6609.

5. in

6.

Novotny, J., Bruccoleri, R. and Saul, antigen-antibody complexes McPC603,

F. A. (1989) On the attribution of binding energy D1.3 and HyHEL-5. Biochem. 28:4735.

Johnston, M. F. M., Barisas, B. G. and Sturtevant, J. M. (1974) Thermodynamics of

hapten binding

to

MOPC 315 and MOPC 460

mouse

myeloma proteins.

Biochem. 13:390.

7. Dower S. K., Wain-Hobson S., Gettins P., Givol D., Jackson W. R. C, Perkins S. J., Sunderland C. ., Sutton B. J., Wright C. E. and Dwek R. A. (1977) The combining site of the dinitrophenyl-binding immunoglobulin A myeloma protein MOPC315. Biochem. J.

165:207.

Barisas, B. G., Singer, S. J. and Sturtevant, J. M. (1972) Thermodynamics of 2,4-dinitrophenyl and 2,4,6-trinitrophenyl haptens to the homologous and heterologous rabbit antibodies. Biochem. 11:2741. 8. of

9.

the

binding

Kussie, P. H., Sherman, . ., Marchetti, D. and Linthicum, D. S. (1989) Molecular

analysis

of monoclonal

idiotypes

and

anti-idiotypes.

10. Scatchard, G. (1949) Attractions of Acad. Sci. 51:660.

proteins

Meth.

Enzymol. 178:917.

for small molecules and ions. Ann.

. Y.

11. Stinson, R.A and Holbrook, J.J. (1973) Equilibrium binding of nicotinamide nucleotides to lactate dehydrogenese. Biochem. J. 131:719. 12.

Froese, A. and Sehon, A. H. (1975) Kinetics of antibody-hapten reactions. Contemp.

Top. Mol. Immunol. 4:23.

13. Beatty, J. W. and Scamehorn, R. G. (1977) Reaction rate theories. In Kinetics, p.77, University Press of America, Washington, D.C. 14. Mukkur, T. K. S. (1980) Biochem. Sci. 5:72.

15.

Thermodynamics

of

:

Reaction

hapten-antibody interaction(s).

Trends

Green, F.A. (1982) Erythrocyte membrane phosphatidylcholine and Rh(D) antigen

cryolatency.

Immunol. Comm.

11:25.

16. Absolom, D. R. and van Oss, C. J. (1986) The nature of the antigen-antibody bond and the factors affecting its association and dissociation. CRC Crit. Rev. Immunol. 6: 1.

309

17. LeNeveu, D.M. and Rand, R.P. (1977) Measurement and modification of forces between lecithin bilayers. Biophys. J. 18:209.

18.

Kussie, P. H., Anchin, J. M., Subramaniam, S., Glasel, J. and Linthicum, D. S. (1991)

Analysis of the binding site architecture of monoclonal antibodies to morphine by using competitive ligand binding and molecular modelling. J. Immunol. 146:4248. 19. for

Cook, C. E. and Drayer, D. E. (1988) Antibodies:

pharmacological

a

studies. Trends Pharm Sci. 9:373.

rich

source

of novel chemical agents

Direct correspondence to: Dr. D. Scott Linthicum Dept. Vet. Pathobiol. College of Vet. Med. Texas A&M University College Station, TX 77843-4467 Received for publication: 11/4/91 after revision: 2/21/92

Accepted

310