DISCUSSION

The only clear interpretation from the tertiary structures of the two peptides

listed (Figs 2.27 and 2.31) are that there is a spatial separation of the residues

grouped into polar and hydrophobic residues (coloured blue and red

respectively, whenever represented). This is important for lipid induced

structures as it provides a basis for interaction between the hydrophobic acyl

chains and the charged or polar surface of the membrane. Since the primary

structure of the tachykinins exhibit the hydrophobic residues in the common C­

terminal regions, it is assumed that this would have a binding site within the

membrane, and so would be presented to the receptor in such a form by the

lipid of the membrane.

Our data for NKA and Kassinin in d-OPC however drives home the point that

structure determination from NMR is an i II-posed 1 problem (Sherman and

Johnson, 1993). Linear peptides are flexible - a point consistently stressed in

this thesis. This inherent flexibility increases the error- associated with

. approximations and assumptions made with the use to the ISPA in analysing

NOE data, to the extent that the structure determined by its use would not be

representative of the true nature of the the peptides conformation. The

discussion builds on data available to describe flexibility in peptides and to

derive information on the conformational preference of the peptides from the

results listed above. We further present a protocol to study flexible peptides

using NMR and end this discussion with information important fo( the bindng

of the tachykinin peptides.

, The mathematical definition of a "well-posed" problem is given as "The problem of determining a solution z=R(u) in a metric space Z from initial data in metric space U is said to be "well-posed" on (Z,U) if (i) for every u in U there exists a solution z in Z;(ii) the solution is uniquely determined; and (iii) the problem is stable" problems for which at least one of the conditions above is violated are termed "ill-posed". (Sherman and Johnson, 1993)

120

Flexibility in peptides:2

Rotations about bonds are described as torsion or dihedral angles,

which are usually described to lie in the range -180 0 to + 180 o. Rotation

about the N-Ca bond of the peptide backbone is denoted by the torsion angle

$, rotation about the Ca - C' bond by \V, and that about the peptide bond (C'­

N) by ro. The maximum value of 180 0 (which is the same as -180 0) is given

to each of the torsion angles in the maximally extended chain, when the N, Ca

and C' atoms are all trans to each other. At the other extreme, the cis

conformation, the rotation angles have the value of zero.

Torsion angles of the side chain are designate by Xi where j is the

number of the bond counting outward from the Ca atom of the main chains.

Rotations about single bonds are intrinsically equivalent, and the

relative preference for each particularly torsion angle is determined by the

energies of the non covalent interactions among the atoms and of the atoms

with their environment. In the case of the peptide bond, the trans form (ro =

0°) is predominant because in the cis form the Ca atoms and the side chains

of neighbouring residues are in too close proximity. The peptide bond also has

a partial double bond characted which maintains it in this form.

When residue I . + 1 is Pro, however, there is very little difference

between the two forms, and the trans form is only slightly favoured, generally

by a ratio of about 4: 1. The peptide bond preceding a Pro residue does not

have the double-bond character that specifies the planar form. Small

deviations from planarity of either the cis or trans form, with Am = -20° to

+10°, are thought to be only slightly unfavourable energetically in most peptide

bonds.

The values of $ and \V that are possible are constrained geometrically

due to steric clashes between non neighboring atoms. The permitted values of

$ and \V were first determined by Ramachandran and colleagues, using hard-

2 All quoted data for this section is from Crieghton, 1991. 71l- {;£72-121

sphere models of the toms and' fixed geometries of the bonds and are

indicated on a two-dimensional map of the cjI-\V plane, in what has come to be

known as a Ramachandran plot. The part of the total area that is fully allowed

is about 7.5% and the part that is partially allowed is 22.5%, which gives a

quantitative measure of the limitations on flexibility of the polypeptide chain.

Gly residues have no en atom and so the restrictions on allowed

conformations are much less severe, and 45% of the total area is fully

allowed, 61% within the extreme limits. In this case, the Ramachandran plot is

symmetric as a re~ult of the absence of chirality of Gly residues.

Additional restrictions arise for the longer, larger side chains of the

other amino acids, and the allowed region is smaller. The Pro residue is a

special case because the relatively rigid five-membered ring drastically limits

the value ot-cjI to approximately -60 0•

More precise Ramachandran plots result from calculations of the

relative energies of each conformation, permitting appropriate flexibility of

bond lengths and angles and evaluating all favourable and unfavourable

interactions, including those with the solvent. The more detailed calculations

indicate much smaller energy differences between the so-called allowed and

disallowed regions than might be expected solely from steric considerations.

Flexibility of bond lengths and angles permits conformations that are not

possible for hard-sphere atoms and rigid bonds, and the variety of interactions

that occur among atoms can cause the. energies to vary substantially.

Nevertheless, the classical Ramachandran plots illustrating the fully and

partially allowed regions are remarkably appropriate for Pl"9teins (Williamson,

1992).

Intrinsic Rates of Bond Rotation in Polypeptides.

The free-energy barriers between the energetically favoured conformations of

a single residue in a polypeptide chain, by rotations of cjI and \V are only of the

order of 0.5-1.5 kcallmol (2-6 kJ/mol), ·so these rotations would be expected to

122

occur at rates of the order 10'2 S-1. Movements in polymers, however, are

complex and not entirely understood. Each conformational parameter of an

ideal random coil can be independent of all others at equilibrium, but this

cannot be the case for fluctuations on a short time scale. If only one bond near

the middle of a chain were to rotate by 180 0, the ends of the chain would have

to undergo extremely large movements, and such a process is implausible in a

viscous solution. It seems intuitively obvious, therefore, that the rotations of

all bonds must be coordinated in such a way as to produce more plausible

types of movement, but a C<?mplete description of how this might occur is not

available.

The average rates at which individual bonds in disordered polypeptides

change conformations are know from their relaxation times measured by 13C

nuclear magnetic resonance. The relaxation time is the average time it takes

a population of molecules to change in some way by e-' of their equilibrium

positions. For a bond rotation, this change is 68°, the angle for which the

cosine has the value e-1. The relaxation times of the en atoms of the

backbone of a disordered polypeptide chan have been measured to be 1.4-2.6

nS,indicating that the bonds of the polypeptide backbone are rotating by more

than 1 ° every 2X 10-11 s. These rotations of the backbone occur more slowly

than they would in a very small molecule; the relaxation time of the side chains

are shorter and decrease for atoms that are farther from the backbone.

The rates at which the ends of a polypeptide chain are moving relative

to each other by diffusion have been measured by· using the efficiency of

fluorescence energy transfer between a fluorescent donor and an acceptor.

Their rates of relative motion were found to be an order of magnitude lower

than the diffusion of a fluorescent donor and acceptor that are not tethered by

the polypeptide chain, indicting that the polypeptide chain possesses

appreciable internal friction that resists motion. Nevertheless, parts of a

disordered polypeptide chain that are separated by 50-100 residues tend to

move through distances that are comparable to their average separation in 10-

5_10-6 s. Therefore, two groups on a disordered polypeptide chain come into

proximity about 105-106 times per second.

123

" ... "

The only conformational transition in disordered polypeptide chains,

known to have an intrinsically high free-energy barrier, is rotation about the

peptide bond, interconverting cis (ro=OD) and trans (ro = 180 0) forms. This

rotation requires disruption of the normal double bonded nature of the peptide,

and the rate of interconversion is not known because the cis form is not

usually populated significantly with a single bond (although the cis population

might be significant when there are many peptide bonds in a molecule). In the

case of Pro residues, however, the situation is different because the cis form

of the preceding peptide bond is of nearly the same e~ergy as the trans form.

This peptide bond might not be expected to have double bond character, due

to the absence of an N-H of the pro residue, but is rate of cis-trans

interconversion is very slow, with a half-time of 20 min at 0 DC. The free­

energy barrier therefore is 20.4 kcallmol. The rate is very temperature

dependent, having an activation· energy of 20kcallmol, so the rate increases

by a factor of 3.3 for each 10 DC rise in temperature within the normal range.

NKA and Kassinin:

In general, an acceptable number of NOEs per residue to determine the

conformation of the peptide or protein is about 11-12 (Williamson, 1992;

Wutrich, 1991).

The Ramachandran plot of the acceptable conformations (acceptable is

defined in terms of NOE violations) for.the two peptides show a large variation

in the $, \II angles for each residue. To explain this variation on the basis of

flexibility, we have simulated all conformations of a tripeptide containing

alanine residues by rotating the $ and \II torsions of the central residue in steps

of 10 degrees, and mapping the energies of resultant conformations. To

describe the relative populations of each conformer, a Boltzmann distribution

was applied to the energy map. Figure (2.31) shows the relative populations of

trialanine. Comparing this figure with our results from the simulated annealing

using NOE restraints, Kassinin (Fig. 2.30) is almost totally flexible, with the

concentration of d-OPC used, while NKA (Fig. 2.26), with a grouping of $, \II

124

torsion values, tends towards some conformational preference - the B-turn

estimated from the NOE patterns can be seen in the overlay of the backbone

atoms for the conformations resulting from the simulated annealing of the

peptide using NOE restraints (Fig. 2.27).

II Developing a protocol for the study of flexible peptide conformation

using NMR

It would be inaccurate to end the discussion at this stage as it would

give the impression that the over-interpretation of NMR data associated by not

incorporating parameters that describe peptide flexibility, is not appreciated. In

this section of the discussion, we present the development of a protocol to

stud peptide structure based on our deductions and clues available from

literature.

The concept of conformational averaging: Small, flexible molecular can exist

in solution in a number of conformations, the population of each weighted with

a Boltzmann distribution, called a conformational ensemble. If conformational

exchange is faster than the NMR time scale (observations are made over a

time scale of milliseconds to seconds), all NMR parameters observed are time

averaged. This can lead to a wrong result, .which may not be representative of

the peptides conformation in nature. as shown in the figure alongside. It is

unfortunate that peptide structure is still analysed using the assumption that

the peptide is rigid, with the statement that "the case presented does not

represent a true structure but a time-averaged approximation to it" (Horne et

aI., 1993).

125

, I

/1 A B

Ll c

Fig 2. F: Confonnational Averaging: In a simple case, if a molecule exists as two confonners with exchange faster than the NMR time scale (represented by A and B), the NOE will average out such that if the ISPA is used, confonner C would result - which is neither representative of the confonnational ensemble, nor of either confonner.

Identifying if multiple conformations exist:

If a peptide is 'rigid' - there would be an overwhelming population of a

·single conformation. This would imply there would be a significant variations

of 3 JNH-u and the temperature dependence of the amides along the sequence.

The clearest indication of rigidity is that all backbone atoms should have the

same correlation time. Increasing use of proton detected 13C and 15N T1

experiments and availability of biosynthetically directed isotope labelling

makes the determination of this approach more feasible.

However this experiment makes a lot of demands on instrument sample

and operator, so the simplest way is to check for averaging among the NMR

parameters. If chemical shifts (especially the Cn protons) assume close to

126

random- coil conformation, J-coupling values show averaged values of around

7 and all the NOEs are not properly described by a single conformation,' a

case of conformational exchange is assumed.

Multiple conformation determination: A review of recent literature

If a peptide populates N structures Si, each with a probability Wi, then

ensemble present in solution is then given by L w.s. where L WI = 1. To

fully describe the ensemble, one needs to thus derive all possible structures.

and their probabilities.

By scanning the available literature, two varieties of multiple

conformational analysis have been used: (1) those that work with probability

distributions of structures and (2) those that work with discrete structures.

(1) a) Noggle and Schirmer attempted to reconstruct the continuous probability

distribution of glycosyl rotamers in nucleotides from NOEs. They fitted

the experimental data to a probability distribution made up of a sum of

Gaussian curves (Schirmer et a!., 1972). This approach has been

difficult to implement and has not gained much popularity.

b) Markley's group measured vicinal homonuclear and heteronuclear spin­

spin coupling constants (J) and NOEs from NMR data. They have

utilised this data to determine the probability distribution of

conformations (a rotamer) by solving for the coefficients of the Fourier

expansion of rotamer probability (Dzakula et ai, 1992 a,b). The amount

of data required for this analysis is however restrictive of the technique

use.

(2).The Monte Carlo procedures of Nikiforovich et al.(Shenderovich et ai,

1984; Nikiforovich et ai, 1987, 1988 and 1993). Ernst's MEDUSA

algorithm (Ernst, 1992; Bruschweiler, 1992) and the Landis' groups

conformer population analysis (CPA) (Landis et ai, 1991, 1995 and

Giovanetti et ai, 1993) all essentially use the same philosophy of

independently generating large numbers of discrete trial structures that

127

are then analysed to' find the combinations of structures that best

satisfy the NMR data. The CPA technique, for instance, comprises

three steps after the collection of 20 NOESY spectra for the molecule

under consideration.

1. The building of trial solution structures using conformational

searching methods combined with molecular mechanics energy

minimisations.

2. Optimisation of rotational correlation times and external relaxation

rate parameters for use in the back-calculation of NOE spectra for

trial solution structures and

3. Conformer population analysis of the observed NOE data, using a

least sequences fit to deconvolute the experimental data as a sum of

the spectra formed from the trial structures.

These techniques are unstable from the point view that they are overly

dependent on the generation of the starting conformation, which may not fully

describe the natural condition.

The time-variable NOE constraint of Torda et al. (Torda et ai, 1989)

combines features of both single and multi-conformational analyses into a time

dependent model that average the NOE as a function of (r\2 a better fit for

fast exchanging conformations. The technique is computationally demanding, . time consuming and hence expensive and has not been applied other than the

authors' initial applications·(Torda et ai, 1990) through Kaptien's group intends

to introduce it towards further improving his IRMA (Iterative Relaxation Matrix

Analysis) program (80nvin et ai, 1993).

The back calculation of NOE spectra to check the validity of the

proposed model of the conformational ensemble is however important.

The Iterative Relaxation matrix Analysis (IRMA)

A matrix containing the direct relaxation rate constants along its

diagonal and the dipolar cross-relaxation rate constants as its off-diagonal

elements is called the relaxation matrix. Since the cross relaxation rates are

proportional to the NOE and correlation time t e , in prinCiple one can simulate

128

the NOE spectrum from a model of 'the molecule. By comparing the

experimental spectrum with that simulated from the model, constraints that

produce a better fit can be generated to build a more appropriate model. The

relaxation matrix analysis has three advantages over strict use of the NOESY

in deriving distance constraints:

(I) it can predict peaks not visible in experiments due to poor signal to noise

resolution,

(ii) it can be used in conjunction with experimental data collected at longer

mixing times, since the initial build up is unnecessary for the calculation

and

(iii) it accounts for spin diffusion.

Kaptien and co-workers have developed one such scheme, which

iteratively uses the relaxation matrix calculations to simulate the spectrum, and

either distance geometry or restrained molecular dynamics to build a model of

the molecule, with fresh constrains generated in comparison with experimental

NOESY spectra (Boelens et aI., 1989). It has evolved, initially using a single

conformer approach to using an ensemble conformations with parameters that

describe dynamics (Koning et al., 1991; Bonvin et al., 1993).

Parameters that describe dynamics: the model-free approach of Lipari

and Szabo.

The most widely used parameters to interpret relaxation data are those

developed in the 'model free' approach of Lipari and Szabo (1982). It is based

on the assumption that the time correlation function can be factored into

contributions from the overall motion and internal motions:

This leads to a spectral density function containing an order parameter S2

(where S2 and 1-S2 are the relative amplitudes of the two correlation functions)

and a correlation time for internal motions "tj.

129

If one assumes that the internal correlatior"l function decays as a single

exponential, one has

........ (3.34)

with

........ (3.35)

The spectral density function. of Eqn 3.34. contains essentially only two fitting

parameters. 52 , the model-independent coefficient for the spectral density

function corresponding to the overall tumbling; and "tl. a model dependent

correlation time for internal motion. The overall correlation time"tc is expected

to be the same for all nuclei of a protein. If"tc is known, measurements of

relaxation parameters, such as T 1 and T 2, are sufficient to determine these

two parameters.

These parameters thus provide an opportunity to determine the

conformation of flexible peptides by incorporating parameters that describe

this flexibility.

For correct usage, an accurate determination of these parameters is

necessary.

Determination of dynamic parameters from computer simulations:

On the theoretical side the internal motions (or the effects of internal

dynamics on the observables) can be calculated from free Molecular

Dynamics Simulations. This idea was first applied to the calculation of

relaxation rates (Levy and Karplus, 1981) and NOE (Olejniczak et ai, 1984) by

Karplus and coworkers. From free MD, the correlation function for each

proton-proton pair involved in an NOE can be calculated (Levy and Keepers,

1986). This method has recently been incorporated into the IRMA program

described above. The theoretical NOEs calculated from the relaxation matrix

are corrected for the internal motions, as ascertained from the free MD. The

130

spectral densities J(ro) are scal~d by the calculated 52 using the Lipari-Szabo

equation described above (80nvin et ai, 1991, 1993).

The addition of solvent in the free molecular dynamics calculation is

essential to prevent the artefacts arising from unreal electrostatic interaction

(Pastor, 1994).

Since our aim is to determine he lipid induced conformation of the

tachykinin, it is necessary to be able to simulate the lipid as part of the solvent

box. In addition to physical conclusions concerning the similarity of bilayer and

hexadecane dynamics, Venable et aI., (1993) reported a novel simulation

methodology: the initial condition for the bilayer were based on independently

generated configurations consistent with a mean field. This has been applied

to simulate gramicidin S within a membrane bilayer (Woolf and Roux, 1995).

To test the simulations, Pastor et aI., (1991) had advocated measuring the

deuterium relaxation times for lipids to compare with simulations. This is a

condition suitable to our experiment procedure as the sample contains

deuterated OPC - to prevent spin diffusion and allow a clear spectra. It may

be an easy step to utilise the information from these experiments as restraints

in the lipid simulation.

Experimental Determination of dynamic parameters:

Classically, 13C T, measurements or proton detected 13C relaxation

parameters (Palmer et al., 1991; Stone et aI., 1992) are used to experimentally

determine the dynamics of a molecule from NMR. Our attempt at performing

this experiment failed, even after deuterating the sample. The sensitivity of the

method may not be enough to probe T 1 relaxation as the concentration of the

peptide is restricted to relatively low levels due to the presence of lipid in the

sample tube.

A more recent technique - the Off-Resonance ROESY could provide an

alternative route to describing the dynamics of peptide (8irlirakis et aI., 1996,

Kuwata and Schleich, 1994). The classical ROESY is complicated by

Hartmann-Hahn transfers and the dependence of cross-peak intensities of the

131

radio-frequency spectrum (81) and resonance offset, which is substantially

reduced in this technique which has part NOESY and part ROESY

characteristics due to the presence of the resonance signal far off-field.

Improved versions use a trapezpoidal spin-lock pulse to improve the

signal to noise ratio and reduce the distortion of antiphase cross-peaks

(Devaux et aI., 1995).

The dipolar spectral density function is the Fourier transform of the

autocorrelation function. The probability of a relaxation transition at a given

frequency is proportional to the spectral density at this frequency. This is the

link between nuclear magnetic relaxation and molecular structure and motion.

Without any assumption about the model of motion, we have:

(J = -J(O) + 6J(2ro)

J.1 = 2J(0) + 3J(ro)

J(O) ~ J(ro) ~ J(2ro) ~ 0

........ (3.36)

........ (3.37)

........ (3.38)

where ro is the Larmor frequency, J(ro) would be the spectral density function

at the lamour frequency ro, and (J and J.1 are the dipolar cross and direct

relaxation rate constants respectively. We cannot obtain the three unknowns

J(O) J(ro) and J(2ro) since there are only two measured quantities (J and J.1.

The last inequalities make it possible to obtain only the upper and lower

uncertainty limits for the spectral density function at three frequencies:

1 1 J (0) = - 2 (J + 4 J.1 + £ ........ (3.39)

J (ro ) = ~ [ (J + ~ J.1 - 2£] ........ (3.40)

1 [1 1 ] J (2ro) = - - (J + - J.1 + £ 624

........ (3.41)

132

........ (3.42)

(Desvaux et ai, 1994)

A possible solution is to study relaxation at two values of the magnetic

field Bo which differ by a factor of two. We then obtain a system of four

equations with four unknowns, and we can sample the spectral density

function at the frequencies 0, 0), 20) and 40) (Devaux et ai, 1995a).

Interproton distances and correlation times

Assuming that the motion of two interacting protons can be described by a

monoexponential autocorrelation function (i.e. their spectral density function is

Lorentzian), the cross relaxation rates are:

.... , ... (3.43)

........ (3.44)

It should be stressed that the assumption made here does not require

equal mobility between different proton pairs, as in a rigid molecule.

Therefore, we can obtain internuclear distances r ( a geometrical parameter)

and local correlation times ( a dynamic parameter) for any pair of protons.

This reduces the intrinsic error resulting from the use of an internal reference

(Desvaux et ai, 1995 a and b).

The above discussion provides us with enough material to suggest a protocol

for the study of flexible peptides. Based on the core of the IRMA programs

(i) free MD is used to provide dynamic parameters ( S2 and tj ) from

computer simulation to scale the NOE data appropriately.

133

(ii) The use of an ensemble of conformations weighted with a Soltzman

distribution is to be used in the above and in calculations involving

the relaxation matrix (van Gunsteren et ai, 1996).

(iii) The experimental determination of order parameters and spectral

density function using off-resonance ROESY, would provide cross­

relaxation information irrespective of the mobility of the molecule.

These values can be used in isolation, as a cross-check on or in

combination with the MD data provided from (i) above.

This protocol is shown in Fig. (2.G)

experimental NOE constraints IRMA

rMD/SA/DG

confonnational ensemble (weighted with a Boltnnann distribution)

use new restraints generated from relaxation matrix

YES

calculate relaxation matrix

Peptide conformational ensemble

FreeMD calculate order parameterss

Off-Resonance ROESY -order parameters

rotational correlation time or directly calculate relaxation matrix

Fig 2.G: Flow chart of protocol outlined in accompanying text to detennine the structure of a molecule more accurately using parameters that describe dynamics and back-calculating the NOE spectrum to compare the results.

134

(iii) Sequence Comparisons Between The Tachykinin Receptors

As is the case with all G-protein-coupled receptors, the greatest

similarities are found in the hydrophobic transmembrane core regions.

Sigemoto et a!. (1990) has shown sequence identities of 66.3 % between NKB

and substance Preceptors, 54.9% between NKB and NKA receptor, and 53.7

% between substance P and NKA receptors. These figures are similar to

those found between different members of the muscarinic or the. p-adrenegic

receptor families. With the amine receptors it is this core domain which is

widely held to contain the agonist binding site, and the acidic residues in TM 2

and TM 3 are potentially important acting as counterions to the positive charge

of the physiological amines. The only charged conserved residues in the

tachykinin receptors are an acidic residue in TM 2 and a basic residue in TM

3. In view of its role in agonist binding in the p-receptor, the acidic residue in

TM2 is likely to play similar role in the tachykinin receptors. However, the

absence of any conserved acidic residue in TM 3 is worth noting, as this plays

a pivotal role in all ligand binding to the P- and muscarinic receptors. Clearly

the model of binding worked out for these small molecules does not apply to

peptides. In this model, the whole ligand binding site is in a pocket formed by

the helices. Studies on 0'2,P2chimeric receptors have shown that TM 6 and

. TM 7 play important roles in discriminating between different ligands. In the

light of this it is interesting to note that the degree of similarity varies between

the different helices. There is almost complete homology in TM7, with over 50

% in helices TM 2, 3, and 6. This leaves most variety in TM 1 and 5. How

this relates to receptor function is unclear. Shigemoto et a!. (1990) have

suggested that the conserved C-terminus of the tachykinin peptides may

interact with TM 7. The size of peptides means that they cannot be entirely

accommodated within the interhelical pocket, and so the extracellular portions

of the tachykinin receptor deserve attention. The interhelical loops are short

and each receptor has a unique pattern of charged acidic and basic residues,

especially loops c3 and 04.

The evidence from endogenous receptors indicates that all three couple

to inositol lipid hydrolysis by phospholipase C, via a pertussis toxin-insensitive

G-protein, leading to production of inositol 1,4,5-triphosphate and

diacylglycerol. This shared mechanism fits with the considerable sequence

resemblance on the cytoplasmic segments of the receptors. Loop i 1 has

almost total conservation, loop i2 has about 75% sequence identity, and there

are considerable sequence similarities in the proximal and distal ends of the i3

loop near the transmembrane segments. It has been suggested that the

proximal and distal ends of this loop are particularly important in conferring G­

protein specificity (O'Dowd et aL, 1988). Another region similarly implicated is

the proximal (membrane) end of the C-terminus, and again this shows a high

degree of conservation among the tachykinin receptor subtypes. The main

differences between the tachykinin receptor subtypes in functional terms is

their rate of desensitisation: SP > NKB > NKA (Shigemoto et aL, 1990). The

C-terminus in the J3-receptor has been shown to be . of importance in

desensitisation (Cheung et aL, 1989), since it is the site at which the kinase

specific for agonist-occupied receptor, {3-adrenergic receptor kinase (J3-ARK),

appears to inactivate normal receptor function (Bouvier et aL, 1988). There is

clear divergence in the sequence of the C-terminal tails of the tachykinin

receptors. Shigemoto et aI., (1990) have drawn attention to the different

number and distribution of serine and threonine residues in the C-terminus

among the three subtypes (26 in the substance Preceptor, 28 in the

substance K receptor, and 14 in the neuromedin K receptor). In addition,

there are differem patterns of serine/threonine residues in the i3 loop. As

kinases of several types are now held to be the final common pathway of

desensitisatio-n (Sibley et aL, 1988), these F9ij" 19S-ffiaybe points at which

regulatory kinases act. In these two regions, the substance P and NKB

receptors have much greater similarity compared to each other than compared

.to f)JKA receptor. The size of the C-terminal tails, especially for the NKB

receptor, raises questions as to whether the tails may have additional

functions apart from G-protein coupling and control of desensitisation, perhaps

in the recognition of other proteins associated with signalling.

Gether et a, (1995) have utilised a different strategy to elucidate the

binding sites and mode of action of tachykinin non-peptide antagonists. They

have used chimeric receptors (by fusing the portions of the genes of the two

receptors corresponding to different sections of the receptor) have attempted

to identify the residues important for tachykinin activity (Gether et al 1993 a,b).

Further results from this kind of experiment will be useful in understanding the

molecular basis of cross-reactivity in the tachykinins.