Localization of agonist and antagonist binding domains of the human ...

Communication

THEJOURNAL OF BIOLOGICAL CHEMISTRY Vol. 267, No. 36 Issue of December 25 pp. 25664-25667 1992 0 1992 hy The American Societ; for Biochemistry and Molecular Biolo& Inc. Printed in U.S.A.

Localization of Agonist and Antagonist Binding Domains of the Human Neurokinin- 1 Receptor*

neurokinin-1 receptor (NKlR) is SP > NKA > NKB, while the rank ordersof potency for the othertwo subtypes (NKBR and NK3R) are NKA> NKB > SP and NKB> NKA > SP, respectively. Previous studies have led to a message-address hypothesis for peptide-receptor interaction. This hypothesis envisions (Received for publication, September 3, 1992) the conserved C-terminal portion of the peptides as an activation message that can be recognized by all three receptor Tung Ming FongS,Ruey-Ruey C. Huang, and Catherine D. Strader subtypes, whereas the divergent N-terminal portion of the peptides acts as a recognition address to determine receptor From the Departmentof Molecular Pharmacology ana’ Biochemistry, Merck Research Laboratories, subtype selectivity (6-8). Analysis of the aminoacid sequences Rahway, New Jersey 07065 of thethreeNKRsubtypeshas revealedahigh level of sequence similarity within the transmembrane domains, while To identify the molecular determinants of ligand- the extracellular and cytoplasmic loops are more divergent (9, receptor interactions, the extracellular domain of the 10). Therefore,itcan be postulated that the commonChuman neurokinin-1 receptor was systematically sub- terminal half of the peptides interacts with the conserved stituted with the corresponding sequences from the transmembrane domains of the receptors, while the unique other two neurokinin receptor subtypes. Three resi- N-terminal half of the peptidesrecognizes the more divergent dues within the first extracellular segment and 2 resi- extracellular domains. dues of the second segment are required for the optimal Studies on other G-protein coupled receptors whose endogbinding of all three natural peptide agonists. The dienous agonists are small molecules have demonstrated that vergent nature of 4 of the 5 residues supports the hypothesis that the peptide binding site on the neuro- the binding sites for theseligandsare located within the kinin-1 receptor is not highly conserved in the other transmembrane domains of the receptors (11-13). Because larger than these non-peptide two receptor subtypes. In contrast, substitution of part the neurokinin peptides are ligands, it seems likely that the extracellular domainsof the of the third extracellular segment and the fourth extracellular segment with the corresponding amino N K l R might comprise part of the ligand binding site. We by systematically substituting the acids of the human neurokinin-3 receptor results inan have tested this hypothesis increase in neurokinin B affinity without affecting extracellular segments of the human NKlR with the corresubstance P binding, suggesting that the two peptides sponding sequence from the human NK3R. If the divergent do not interact with the same set of functional groups extracellular sequences of the receptors determine the rank on the receptor. Among the four extracellular regions, order of potency of peptide agonists, then substitutionof the only parts of the third and fourth segments affect the extracellular sequences in the NKlR by the homologous sebinding of the quinuclidine antagonist L-703,606, and quences from theNK3Rshould cause anincrease in the these two regions may partially account for the neu- affinity of NKB with a concomitant decrease in SP affinity. rokinin- 1 receptor subtype specificity of this non-pep- The present results confirm the contribution of the extraceltide antagonist. These studies demonstrate that both lular domains of the NKlR topeptide binding. However, the the extracellular and transmembrane domains of the data also suggest that the two peptides do not necessarily neurokinin-1 receptor are involved in the binding of interact with the same setof residues on the NKlR. Theresubstance P and related peptides. fore, the peptide binding domains in the neurokinin receptors are more complicated than what the simple message-address model would predict. The peptide neurotransmitters substance P (SP),’ neuroEXPERIMENTALPROCEDURES kinin A (NKA), and neurokinin B (NKB) are characterized All mutant receptors were constructed from the human NKlR by by the common C-terminal sequence FXGLM-NH2. The bieither the polymerase chain reaction method (Perkin-Elmer Cetus) ological actions of neurokinins are mediated by three subtypes or the uracil substitution method of site-directed mutagenesis (Bioof the neurokinin receptor (NKR). These receptors are mem- Rad). All mutated sequences and any sequence that was derived from bers of the G-protein coupled receptor family that is charac- polymerase chain reaction were confirmed by DNA sequencing terized by seven putative transmembrane helices (1-5). The (United States Biochemical Corp.). All receptors were expressed in COS cells to determine the ligand binding affinity. Some mutant rank order of potency of the neurokinin agonists for the * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “aduertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. $ To whom correspondence should be addressed Dept. of Molecular Pharmacology and Biochemistry, 80 “213, Merck Research Laboratories, P. 0. Box 2000, Rahway, NJ 07065. Tel.: 908-594-6711; Fax: 908-594-3337. ‘The abbreviations used are: SP, substance P; BHE, BoltonHunter labeled eledoisin; BHSP, Bolton-Hunter labeled substance P NK, neurokinin; NKlR, neurokinin-1 receptor; NKA, neurokinin A NKB, neurokinin B.

receptors were also expressed in Xenopus oocytes to determine the functional activity (10, 14). The binding affinities of various ligands for the NKlR and its mutants were determined using ‘251-Bolton-Hunter labeled S P is an iodinated analog of the NK1(BHSP) or [’251]L-703,606, which specific antagonist CP-96,345 (15, 16), in the presence of varying concentration of unlabeled ligands. lZ5I-BHSPwas used when the Kd value of SP was smaller than 10 nM, [‘251]L-703,606 was used when the Kd value of S P was larger than 10 nM, and ‘251-Bolton-Hunter labeled eledoisin (BHE) was used for the human NK3R and the E3 mutant of the human NKlR. Thefinal concentrationof radiolabeled ligand was0.2nM. Intact COS cells were used in the ‘“I-BHSP or T - B H E binding assay, whereas plasma membranes were used in the

25664

SiteBinding Ligand ['251]L-703,606 bindingassay. In the case of '"I-BHSP, the data were fitted to the equation (cpm(L) - cpm(1 p M SP))/(cpm(O)- cpm(1 p~ SP)) = ICIo/(L + ICao),in which cpm(L) and cpm(0) represent bound lZ5I-BHSPin the presence andthe absence of unlabeled ligand, respectively, L represents the concentration of unlabeled ligand, and IC5o represents the concentration of unlabeled ligandthat causes 50% inhibition of the specifically bound '"I-BHSP. In the case of [lZ5I]L703,606 binding in the presence of unlabeled L-703,606,the data were fitted as described above, Inthe case of ['251]L-703,606 binding in the presence of agonist, the data were fitted to the equation (cpm(L) cpm(1 p~ L-703,606))/(cpm(O)- cpm(1 p M L-703,606)) = F X K H / ( L + KH) + (1 - F ) X KL/(L+KL)in which F is the fraction of NKlR in the high affinity state, KH is the agonistaffinityfor the high affinity state, and K L is the agonist affinity for the low affinity state. The IC5ovalue was then solved numerically from the fitted curve.

of NK1R El

25665 E2

E3

E4

RESULTS

To probe the role of the extracellular loops of the NK receptors in agonist and antagonist binding, theseregions of the human NKlR were systematically replaced with the analogous regions from the human NK3R (Fig. 1). The resulting mutant receptors were transiently expressed in COS cells, and their ability to bind peptide agonists anda non-peptide quinuclidineantagonist wasassessed. Substitution of the second extracellular loop of the NKlR (E2, residues96-108) FIG. 1. Model of the N K l R along with sequence alignment with the analogous region of the NK3R resulted in undetect- of the human NKlR, NKSR, and NK3R in the extracellular domains. The amino acids of the NKlR are in circles or squares, able binding of lZ5I-BHSP at 0.2 nM. In contrast, the nonwith the amino acids of the NK2R listed on the left and the amino peptide antagonist ['251]L-703,606bound to both theE2 mu- acids of the NK3R listed on the right. Amino acids conserved among tant and thewild type N K l R with the same affinity(Fig. 2). all three subtypes are indicated by a single letter. Filled circles repreT h e binding affinitiesof both SP and NKBfor the E2 mutant sent conserved residues, and open circles represent divergent residues. were greatly reduced (Fig. 2). However, G-protein activation Squares indicate that replacing 1residue affects peptide binding. The by the E2 mutant was normal. InXenopus oocytes expressing numbering is based on the human NKlR, and only the sequence from the E2 mutant, agonists elicited an oscillating calcium-acti- 21 to 313 is presented. vated chloride current that is characteristic of activation of 0 L-703,606 V SP 0 NKE the phospholipase C-mediated phosphatidylinositol pathway (14).As would be expected from the reduced binding affinities of agonists for the E2 mutant receptor, the dose-response curves for SP and NKBwere shifted to the right compared to the wild type receptor (Fig. 3B). Substitution of the entire third extracellular domain (E3) of the NKlR with the correspondingsequence of the NK3R resulted in a mutant receptor with no detectable binding of lZ5I-BHSP or['251]L"703,606.In addition, the E3 mutant did not bind lZ5I-BHE, which has high affinity for the NK3R. Because the separationof free and boundradioactive ligands by filtration is possible if the dissociation rate constant is smaller than0.5 s" (equivalent toK d < 20 nM), a lower limit of IC,, > 20 nM was assigned to the E3 mutant for ligands. all However, the E3 mutantreceptor was fully functionalin Xenopus oocytes (Fig. 3C), indicating that the receptor was correctly processed. The EC50 values for the E3 mutant were 20 also consistent with reduced peptide binding affinities compared to thewild type. Three smaller substitutions in theE3 region were constructed in order to analyze the contribution 0 -1.710-0 -8 -7 -6 -5 -1*10-0-1-7-6-, of this extracellular loop to ligand binding. Substitution of log [Ligand] , M residues 170-174 by the homologous NK3R residues (E3a mutant) resulted in a change in the rank order of agonist FIG. 2. Binding affinity measurements for the human potency,with S P > NKB > NKA(Table I; Fig. 2). The N K l R , the human NK3R, and five substitution mutants of the affinity of SP for the E3a mutant was similar to that of the human N K l R . The human NKlR is shown using both lZ5I-BHSP wild type NKlR.However, the affinityfor NKB was increased or ['251]L-703,606(top panels). Each datapoint is the average of Each curue is representative of at least 2 similar experi3-f0ld,suggesting that this region of the NK3R might be duplicates. ments. important for NKB binding. A second substitution in this region, in which residues 176-183 were replaced withthe the binding affinities for the three peptideswere not signifiNK3R sequence (E3b mutant), resulted ina reduction in the cantly affected. In contrast, the binding affinity of the antagbindingaffinities for all three peptide agonists, while the onist L-703,606 was reduced 200-fold (Table I; Fig. 2). In the affinity for the antagonist L-703,606 was not affected (Table wild type NK3R, theKd of L-703,606 was greater than 1 KM. I; Fig. 2). Finally, when residues 187-195 of the NKlR were Substitution of the fourth extracellular loop of the NKlR substituted with theanalogous NK3R residues (E3c mutant), (E4 mutant,residues 271-280) with the corresponding NK3R

I't -

Ligand BindingSite of N K l R

25666

sequencedidnot significantlyaffect SP bindingaffinity (Table I). However, both NKB and NKA exhibited 5-8-fold increases in binding affinity for the E4 mutant compared to the wild type NKlR. A parallel decrease in affinity of the NK1 selective antagonist L-703,606 was observedfor this mutant receptor(Fig. 2). In addition, a point mutation in helix 7 was constructed to replace methionine 291 with the NK3R homolog phenylalanine. A %fold increase in NKB affinity was also observed for the M291F mutant (Table I). The systematic replacement of the extracellular loops of t h e N K l R describedabove implicatestheE2 loop in the binding of all three peptideagonists. Our previous analysis of the N-terminal domain (El region) also indicated that residues 21-29 are important for high affinity peptide binding (17). To elucidate the precise role of individual residues in thesesegmentsinpeptidebinding,pointmutations were analyzed. The three non-conserved residues at positions 21, 22, and 29 in the E l region of the human NKlR can be substitutedwiththecorrespondingaminoacids from the NK3R without any effect on agonist or antagonist binding

-

(Table I). Likewise, substitution of the conserved proline 28 and tryptophan 30 with alanine did not affect the ligand interactions. In contrast, substitutionof the conserved phenylalanine 25 with alanine resulted in a dramatic decrease in affinity for all three neurokinin peptides. L-703,606 binding was notaffected by thissubstitution. Asparagine23 and glutamine 24 are conserved between the NKlR and NK3R, while the NK2R contains threonine and alanine, respectively. Substitution of asparagine 23 and glutamine 24 also resulted in a large reduction in the affinities of all three peptides for the NKlR. These substitutions did notaffect the binding of the antagonist L-703,606. In the second extracellular loop, 6 residues are divergent between the NKlR and NK3R (Fig. 1).Substitution of the non-conserved asparagine96 and histidine 108 with their NK3R homologs (serine and glutamine, respectively) resulted in a substantial reduction in theaffinities forall three agonist peptides, although the antagonist affinity was not affected by these point mutations (TableI). DISCUSSION

m

The present studywas designed to identify the contribution of the extracellular domains of the NKlR to agonist and antagonist binding. None of the extracellular loop replacements describedhere resulted inareversal in the relative affinities of SP and NKB. Therefore it isprobable that some residues in the transmembrane domain of the receptor may confer SP specificity on the NKlR. On the other hand, the -10-9 -8 -7 - 6 -5 present results indicate thatseveral residues in the N-termiLT log [Peptide] , M nal domain (N23, Q24, F25), loop E2 (N96, H108),and loop E3 (176-183) are required for the high affinity binding of all FIG.3. Agonist-elicited chloride current response in Xenopus oocytes expressing the wild type human N K l R ( A ) , t h e three peptides. However, these mutations do not affect the E2 m u t a n t ( B ) ,or t h e E3 m u t a n t (C). All oocytes were injected rank order of potency of the three neurokinin peptide agowith 2 ng of in vitro transcribed RNA. Each point is the average of nists, and they are not required for the binding of the NK1a t least three measurements. The data are expressed as percent of selective antagonist L-703,606. The absence of a decrease in the maximal response elicited by SP. For the wild type, SP at30 nM the affinityof the antagonist argues against effect any of these elicited a maximal response of 1768 f 245 (7) nA. For the E2 mutant, SP at 3 p~ elicited a maximal response of 2530 f 236 (4) nA. For the mutations on the overall conformation of the receptor, alE3 mutant, SP at 3 p M elicited a maximal response of 1196 f 261 (4) though local conformational effects within the loops cannot be determined a t present. The data suggest that theseresidues nA. o SP

0

NKB

TABLEI Binding affinity measurements of the wild type human NKlR, 16 mutants of the human NKIR,and the wild type human NK3R Mean f S.E. are listed. ICbo b M ) Receptors

SP

0.63 f 0.15 (7) 3.9 f 0.7 (4) 6,000 f 2,000 (2) >20 1 (2) 470 f 250 (2) 1.3 f 0.6 (4) 1.4 (2) 42 f 6 (3) 0.45 f 0.05 (2) 750 f 250 (2) 90 f 10 (2) Q24A: F25A >10,000 (2) 0.45 +. 0.05 (2) P28A" 0.45 f 0.05 (2) W30A" >10,000 (2) FPW(25,28,30)AAAb 180 _t 22 (2) N96S' 150 f 50 (2) H108Q' M291F" 0.3 (2) Data are determined using IZ5I-BHSP. 'Data are determined using ['251]L-703,606. ND, not determined. Data are determinedusing lZ5I-BHE. hNKlR wild type" hNKlR wild type' E2 = (96-108)NK3' E3 = (170-195)NK3",b , d E3a = (170-174)NK3" E3b = (176-183)NK3' E3c = (187-195)NK3" E4 = (271-280)NK3" hNK3R wild typed EPA(21,22,29)LNSn N23Tb

NKA

NKB

L-703,606

31 f 5 (5) 89 f 31 (3) ND' >20 63 f 34 (2) 500 ? 30 (2) 67 -C 16 (4) 6 k 1.5 (3) 19 f 4 (3) 20 f 1 (2) >10,000 (2) 5,000 k 120 (2) >10,000 (2) 36 f 4 (2) 33 f 12 (2) ND >1,000 (2) >1,000 (2) 14 +. 1.5 (2)

81 f 13 (5) 370 f 106 (3) >10,000 (2) >20 25 f 5 (2) 10,000 f 700 (2) 130 f 31 (4) 11 f 1.5 (3) 2.3 f 1.3 (3) 64 f 27 (3) >10,000 (2) 5,700 f 600 (2) >10,000 (2) 62 f 23 (3) 105 & 4 (2) ND >1,000 (2) >10,000 (2) 30 f 2 (2)

1.4 & 0.5 (2) 0.3 (2) 0.6 f 0.2 (2) >20 1.2 f 0.4 (2) 1.4 & 0.6 (2) 200 (2) 8.5 f 0.5 (2) >1,000 (2) ND ND 0.4 & 0.1 (2) 0.45 (2) ND ND 0.4 0.2 (2) 0.3 (2) 0.5 (2) 1 (2)

Ligand Binding Site of N K l R

25667

of the mutations analyzed here affected the intrinsic activain the external loops of the NKlR may interact with the conserved C-terminal half of the neurokinin peptides, or theytion of theintracellular effector. By analogy withthe (3residues in may be required to maintain the local conformation of the adrenergicreceptor,onemightspeculatethat peptide binding site. The divergent nature of these residues transmembrane helices 5 and 6 and their connecting third implies that the peptide binding sites of different receptor intracellular loop might be involved in agonist-medicated Gprotein activation.The present study supports the hypothesis subtypes may not be very conserved. Substitution of residues 170-174 in E3 and 271-280 in E4 that the structures of the peptide binding sites differ for the by the corresponding NK3R sequence increases the affinity three neurokinin receptor subtypes. of the NKlR for NKB. However, the SP binding affinity is not significantly affected by thesesubstitutions, suggesting Acknowledgments-we thank Dr. D. Burns and €3. Frances for that divergent residues in ~3 and ~4 affect NKB binding but synthesizing [1251]L-703,606,Dr. M.A. Cascieri for critically reading the manuscript, and H. Yu for help with tissue culture. not SP. Factors other than amino acid side chains may also contribute to ligand binding specificity, as it was demonNote Added in Proof-Some of the datawere presented in abstract strated in the arabinose binding Protein where substrate spec- form (26). After the submission of the present report, another abstract ificity and affinity are determined by bound water molecules appeared that described the binding properties of chimeric mutants (18).The present observations imply that different peptides between the rat NKlR and NK3R (27). may not interact with the same set of functional groups on REFERENCES the and thereforethe active conformations Of differ1. Iversen, L. L., Watling,K.J.,McKnight, A. C., Williams, B. J.,and Lee, e n t neurokinin peptides may be different. Consistent with C. M. (1987) Top. Med. Chem. 6 5 , 1-9. 2. Mag& ,' J. E. (1988) Annu. Reu. N e u r o s c ~1 1 , 12-28 this hypothesis are NMR experiments demonstrating that 3. Regoli, D., Drapeau, G., Dion, S., and D'Orleans-Juste, P. (1989) P h a r m different neurokinin peptides adopt different conformations cology 3 8 , 1-15 in solution (19, 20). Analysisof conformationally constrained 4. Helke, c. J., Krause, J. E., Mantyh, P. W., Couture, R., and Bannon, M. J. (1990) FASEB J. 4,1606-1615 SP analogs has also suggested that different peptide confor5. Nakanishi, S. (1991) Annu. Reu. Neurosci. 14,123-136 mationsare recognized by differentNKRsubtypes(21). 6. Schwyzer,R. (1987) EMBO J. 6,2255-2259 7. Buck, S. H., Pruss,R. M., Krstenansky, J. L., Robinson, P. J.,and Taken together, the results presented above indicate that Stauderman, K. A. (1988) Trends Pharrnacol. Sci. 9,3-5 peptide binding involves regions within all four of the extra8. Munekata,E. (1991) c o w . €&hem. PhYsiOl. 9 8 c , 171-179 9. Takanashi, K., Tanaka, A,, Hara, M., and Nakanishi, S. (1992) Eur. J. cellular domains of the NKlR. The transmembrane domain Biochem. 204,1025-1033 is still important for peptide binding as illustrated by the 10. Huang, R. R. C., Cheung, A. H.,Mazina, K. E., Strader, C. D., and Fong, T. M. (1992) Biochem. Biophys. Res. Commun. 184,966-972 point mutation M291F. In contrast, substitutions in the E3c 11. Khorana, H. G. (1992) J . B i d . Chern. 2 6 7 , 1-4 and E4 region reduce the binding affinityof the quinuclidine 12. Strader, c. D., sigal, I. s.,and Dixon, R. A. F. (1989) FASEB J . 3, 18251832 antagonist L-703,606 without affecting SP binding. This is 13. Wheatley, M., Hulme, E. C., Birdsall, N. J. M., Curtis, C. A. M., Eveleigh, P., Pedder, E. K., and Poyner, D. (1988) Trends Pharmacol. Sci. Feh. different from the p-adrenergic receptor, where deletion of Suppl., 19-24 residues in the extracellular segments did not affect the bind14. Fong, T. M., Anderson, S. A,, y u , H., Huang, R.-R. c., and Strader, C. D. (1992) Mol. Pharmacol. 41,24-30 ing Of molecule ligands (22)' Overlapping but non- 15. Snider, R. M., Constantine, J. W., Lowe, J. A,, 111, Longo, K. P., Lebel, W. identical agonist and antagonist binding sites have been des.,Woody,H. A., Drozda, s. E.,Desai, M. C., Vinick, F. J., Spencer, R. W., and Hess, H. J. (1991) Science 251,435-437 T. M., Sadowski, S., Bansal, A,, Swain, C., Seward, E., Frances, B., Burns, D., andStrader, C. D. (1992) Mol. Pharrnacol. 42,458-463 Fong, T.M., Yu, H., Huang, R. R. C., andStrader, C.D. (1992) Biochemistry, in press Quiocho, F. A., Wilson, D. K., and b a s , N. K. (1989) Nature 3 4 0 , 404407 Levian-Teilelbaum, D., Kolodny, N., Chorev, M., Selinger, z., andGilon, C. (1989) Biopolymers 28,51-64 Saviano, G., Temussi, P. A,, Motta, A,, Maggi, C. A,, and Rovero, P. (1991) BLochemistry 30,10175-10181 Cascieri, M. A,, Chichi, G. G., Freidinger, R. M., Colton, C. D., Perlow, D. S., Williams, B., Curtis, N. R., McKnight, A. T., Maguire, J. J., Veher, D. F., and Liang, T.(1986) Mol. Pharmacol. 2 9 , 34-38 Dixon,R. A. F., Sigal, I., Candelore, M. R., Register R. B. Scattergood, W., Rands, E., and Strader, C. D. (1987) EMBO J. 6 , 326G-3275 Strader, C. D., Sigal, I. S., Register, R. B., Candelore, M. R. Rands E. and Dixon, R. A. F. (1987) Proc. Natl. Acad. Sci. U. S. A. 84,'4384-438i Kobilka, B. K., Kobilka, T. S., Daniel, K., Regan, J. W., Caron,M. G., and Lefiowitz, R. J. (1988) Science 240,1310-1316 Wadswofih,H. L., Chazenbalk, G. D., Nagayama, Y., Russo, D., and Rapoport, B. (1990) Science 2 4 9 , 1423-1425 Fong, T. M., Yu, H., Huang, R.-R. c.,andStrader, c. D. (1992) soc. Neurosci. Abstr. 1 8 , 454 Gether, U., Johansen, T. E., Snider, R. M., Lowe, J. A. 111 McLean S Nakanishi, S., and Schwartz, T. W. (1992) SOC.Neuroici. Abstr. 18,'406

scribed for the p2-adrenergic receptor and a2-adrenergic receptor (23, 24). Because the specificity of L-703,606for the

16. Cascieri,M. A., Ber, E., Fong,

NKIR subtype be completely accounted for by the substitution inE3c and E4regions, we hypothesize that amino acid residues within the transmembrane domainof the NKlR are also involved in the binding Of this quinuclidine antagoFurther studies be required to identify specific residuesthatinteract directly withfunctional groups onthe agonist or antagonist molecule. The involvement of all four extracellular loops of the NKlR in the binding of peptide agonists permits a relatively large surface area for therecognition of peptides. This is the first demonstration for a Of four extracellular domains in the binding of small peptides to a G-protein coupled receptor. T h e involvement of the extracellular N-terminal residues in the binding of glYCOProtein hormones has also been demonstrated in a subgroup Of the G-protein-coupled receptors that contain a very large N-terminal domain (25). However, none

17. 18. 19. 20.

21. 22. 23. 24. 25. 26.

27.