La confederación Perú – boliviana

Difficulties arise when these techniques are applied to the assignm ent of reso­ nances in th e spectra of larger proteins (M r> 10,000). There are two m ain reasons for this. W ith th e increase in the size of the molecules, there is an increase in the num ber of signals in th e NMR spectrum . This means there is m ore chance of the cross-peaks having th e sam e chemical shift. Consequently it is no longer possible to distinguish between signals from two different protons. The entire spectrum can ap­ pear very crowded, so m uch so th a t it is virtually impossible to m ake unam biguous assignm ents for many resonances. In addition to this overlap problem the increase in m olecular weight gives rise to larger linewidths causing problems for correlation experim ents involving scalar connectivities. In a COSY experim ent th e cross-peaks are antiphase in nature, having a positive peak adjacent to a negative peak. W ith the increased linew idth, these opposing parts can interfere destructively m aking the final signal absent or deceptively weak.

These problem s have been largely solved by th e development of techniques which improve th e resolution of spectra by (i) increasing th e dim ensionality of th e spec­ tru m and (ii) using heteronuclear experim ents w ith isotopically labelled samples.

Addressing the first point, W üth rich ’s assignment procedure relies on using correlation experim ents to identify th e through-bond connectivities, followed by NOESY experim ents to identify th e sequential through-space connectivities along th e polypeptide chain. One of the first applications of increasing th e dim ensionality

of hom onuclear experim ents was th e reported 3D TOCSY-NOESY (O schkinat et

a i 1988; O schkinat et al. 1989). However the applicability of this hom onuclear 3D

experim ent is lim ited as the efficiency of the through-bond proton-proton correla­ tion step is still poor in large macromolecules. There is also a digital resolution problem in th a t so m any increm ents have to be acquired u ntil a suitable spectrum

can be obtained.

Using heteronuclear NM R experim ents the resolution problem can be im proved by separating th e signals on the basis of individual chemical shifts of th e directly bonded heteronuclei. Originally heteronuclear NM R experim ents were little used for studies of proteins because of the low inherent sensitivity, low n a tu ra l ab u n ­ dance and in stru m en t lim itations. However, all of these problems have been largely overcome. One of the m ost im p o rtan t factors in this has been the developm ent of pro to n -d etected heteronuclear experim ents th a t are more sensitive th a n those where th e low gyrom agnetic heteronucleus is detected (inverse detection). Initially, such experim ents could not be perform ed on commercial NM R spectrom eters, b u t now m ost are equipped for these experim ents (W agner 1989).

Isotopic labelling has become easier w ith the use of improved protein expression system s. This usually involves incorporating the appropriate gene into a bacterial cell and th en growing cells in a m edium containing certain labelled m aterial as food sources for bacterial growth. The bacteria will then take up the labelled m aterial. T he labelled protein is obtained following cell lysis, protein isolation and finally purification. T his approach is less expensive th an having to use large qu an tities of labelled am ino acids in th e growth m edium . In order to produce large am ounts of labelled protein, over-expression systems have been developed where the b acterial strains contain a kind of switch (either chemical or physical) to in itia te p ro tein production at a higher ra te w ithin th e cells.

Isotopic enrichm ent has been employed in protein NM R studies in several useful ways. It has already been m entioned th a t labelling can help overcome th e pro b ­

lems of spectral overlap. Incorporation of into th e sam ple will remove signals

can also help in minimizing unw anted spin-difFusion pathw ays such th a t p articu ­ lar NOEs can be found which will provide b e tte r quality distance inform ation for stru c tu re determ ination. Isotope editing and filtering experim ents have also been

proved very useful. For example, after labelling, experim ents can be run in

which the only signals in th e spectrum are those involving protons attach ed to nuclei. This la tte r exam ple is extrem ely im portant in th e use of three-dim ensional techniques. Here the th ird frequency dimension will be th e resonance frequency of th e heteronucleus (^®N). For example, in a 3D NOESY-based heteronuclear ex­ perim ent, the NOE signals observed would be from ^^N-^H protons to any other protons in th e sample. T he spectra consist of a series of 2D -N 0E SY spectra each

corresponding to protons at each p articular ^®N frequency which can be

separately examined.

A first step in this development of heteronuclear techniques was to convert 2D ’relayed’ experim ents into 3D heteronuclear experim ents (2D and 3D HMQC-

NOESY; (Fesik & Zuiderweg 1988; Gronenborn et ai 1989), 2D NOESY-HM QC;

(G ronenborn et al 1989; M arion et ai 1989; Zuiderweg & Fesik 1989). However,

these techniques alone were not sufficient to make sequence-specific assignm ents, since it was not possible to distinguish between intra- and inter-residue cross-peaks. In 1989 the use of both 3D ^H-^^N TOCSY-HM QC and 3D NOESY-HM QC to pro­ vide th e through-bond and through-space connectivities necessary for th e sequential

assignm ent procedure was reported (M arion et ai 1989). Because th e cost of ^^N-

labelling is relatively low, many groups adopted these techniques and th ere are num erous examples in th e literature. Typical examples include Staphylococcal nu­

clease (Torchia et ai 1988), interleukin-1/9 (M arion et al 1989), anphylatoxin C5a

T h e heteronuclear m ethods rely on one or two bond heteronuclear (H, N, C) couplings. These couplings are much larger w ith respect to th e linew idths and th e cross peaks in the scalar experim ents are less sensitive to th e larger linew idths as­ sociated w ith bigger proteins. This overcomes some of th e problem arising from techniques dependent on H-H couplings (COSY and TO CSY ). T he NIH group

(Ikura et ai 1990; Kay et ai 1990) developed experim ents to exploit this factor.

These m ethods are known as double- and triple-resonance experim ents since they give inform ation about two or th ree different nuclei in one experim ent. T he double resonance techniques are the HCCH-COSY and HCCH-TOCSY experim ents (Bax

et ai 19906; Bax et ai 1990a). They both correlate all th e nuclei (in ^^C labelled proteins) in th e sidechains of th e constituent amino acids w ith the corresponding

C a and H a. Thus these experim ents connect th e sidechain to the backbone pro­

tons. These m ethods have been widely used as the first step in protein stru c tu re d eterm ination where it is necessary to assign the spin system s to residue types. Triple resonance experim ents, as th e name suggests, involve and correlate three

different nuclei. For exam ple th e HNCA experim ent (Ikura et a i 1990; Kay et a i

1990) correlates the am ide proton, th e amide nitrogen and th e alpha carbon. Since

th eir first appearance in th e literatu re (Ikura et ai 1990) th ere have been over 20

different types of 3D triple resonance m ethods developed correlating a variety of

backbone (also C(3 and H/?) atom s and providing sequential inform ation.

These experim ents have proved to be im p o rtan t not only for providing sequential inform ation bu t also for spectral simplification. Since m any have been devised

specifically to correlate backbone atom s, there will usually only be 1 - 2 signals per

residue, hence th ere will be fewer cross-peaks throughout th e spectrum . This m eans th a t it takes less tim e to analyze th e spectra and to make th e assignments.

One of th e first proteins to be assigned using double and triple resonance exper­

im ents was calm odulin (16.7kDa), (Ikura et ai 1990). Since this protein is largely

a-helical, th e chemical shift distributions for the H a and NH protons resonate in a very narrow range. It consists of 4 domains w ith substantial sequence homology and this results in m any of the connectivity p attern s being overlapped.

O th er proteins studied with 3D triple resonance techniques include interleukin-

4 (15kDa), (Powers et al. 1992), hum an interferon- 7 (31.4kDa) (Grezsiek et ai

1992), hum an-im m unodehciency-virus- 1 protease (H IV -1) bound to an inhibitor

(99 residues),(Y am azaki et ai 1994), hum an profilin (15kD a),(M etzler et a i 1993)

and hum an strom elysin- 1 (19.5kDa), (Doren et ai 1993).

Several groups have already m ade use of 4D triple resonance experim ents. Interleukin-

1/? (M arion et a i 1989) was one of th e first to be investigated. O ther 4D studies cited

include c-H -ras p21(l-166).G D P complex (Kraulis et al 1994), hum an interferon- 7

(G rezsiek et a i 1992), and a dom ain from the anti-digoxin antibody (C onstantine

et ai 1993).

2.1

Introduction to N M R

In m odern Fourier Transform NMR, one or m ore radio frequency (rf) pulses are used to p e rtu rb the spin system. The response to this p ertu rb atio n is recorded as th e free induction decay (FID). This is th e m easurem ent of the s ta te of the system im m ediately after the pulse (or pulses) and as it retu rn s to its equilibrium state. T hus th e FID is a m easurem ent of how th e pertu rb ed system behaves over a period of tim e, i.e it is a tim e-dom ain representation of w hat is occurring. To convert this into frequency inform ation (the frequency-domain spectrum ) th e d a ta

90

Figure 2.3: A 90° pulse.

are ’Fourier transform ed’. A Fourier transform is a m athem atical function which converts functions varying w ith respect to tim e into functions of frequency.

W hen a sam ple is placed in a m agnetic field, th e nuclei can align them selves in two orientations, either w ith or against the external field. In fact, at th erm al equi­ librium , th ere will be slightly more nuclei aligned w ith the field (low energy state). T his gives rise to a resu ltant m agnetization along th e m agnetic field direction.

D uring an NM R experim ent this resultant m agnetization can be flipped into different orientations by applying th e appropriate rf pulse. By convention, the equilibrium m agnetization generated by th e nuclei aligned w ith th e field is said to lie on the Z axis in cartesian space and are tipped by rf pulses into th e XY plane. For exam ple a 90° rf pulse applied along the X axis will tip the Z m agnetization into the XY planeialong the Y axis (figure 2.3).

A pulse sequence usually begins w ith a 90° rf pulse of a given phase (X or Y) which tips th e m agnetization into the XY plane Once in this transverse plane, the m agnetization can be m anipulated in m any ways. Subsequent pulses used in the sequence depends upon how th e spin system is to be studied. T he transverse mag­ netization can be detected directly to give a conventional one dim ensional NM R

sp ectru m or it can be m anipulated by further rf pulses if the object is to stu d y or m ake use of interactions between different spins in th e sample. T he exact config­ u ratio n of pulses (the pulse sequence) is chosen according to how the system is to be probed. Several sim ple pulse sequences have been developed for extracting vari­ ous types of inform ation about th e nuclear spins in the sample under investigation. A lthough a large num ber of sequences have been developed, they are usually based on a relatively sm all num ber of sequences th a t can be regarded as building blocks for th e m ore complex sequences. In th e following section I will discuss the building blocks which are constituents of th e pulse sequences used in this study.

2.1.1

T w o- and T hree-D im ensional E xperim ents

A general pulse sequence for a two-dimensional (2D) experim ent can be represented cLS shown in figure 2.4.

At the beginning of th e experim ent the system is in equilibrium with th e mag­ netizatio n vector aligned along th e Z axis. The 90°% pulse tips this longitudinal m agnetization into th e XY plane (transverse m agnetization) along th e Y axis.

D uring th e evolution period, conventionally term ed ti, the m agnetization vectors

90

t l

_mixing

---►

period

evolution

begin to process at their Larmor frequencies and at the end of ti these vectors will have evolved through an angle th a t depends on their frequencies and th e tim e ti.

For exam ple, if we have a proton th a t has a frequency of it will have evolved

through by th e end of th e evolution period, this is known as frequency labelling.

N ote th a t it is the m agnetization which becomes frequency labelled.

T h e object of the ’m ixing’ period (c - d), is to allow some ty p e of m agnetization transfer to occur. The specific n atu re of the pulses and delays used in th e m ixing period will depend upon which type of inform ation one requires, (i.e. coupling inform ation or distance inform ation). During the mixing period some m agnetization will be transferred from one spin to another, so th a t, for exam ple, some transverse

m agnetization of a spin m ight be generated from the m agnetization which

was present at the sta rt of the mixing period. The am ount of transfer depends th e

length of th e mixing period and the strength of th e interaction.

D uring the detection period t2, the transverse m agnetization created by th e

m ixing process undergoes free precession as it did during ti , however now th e free induction decay is recorded. The signal from the free induction decay is a function

of th e real tim e variable t2 and also the evolution tim e t%. In order to ob tain

the two-dim ensional frequency dom ain spectrum , a set of FIDs are acquired w ith different ti values and th e resulting d a ta are Fourier transform ed w ith respect to ti

and t2- Two types of signal appear in the homonuclear 2D spectrum , cross-peaks

and diagonal peaks. As has been previously said, during ti, th e m agnetization

processes at the Larmor frequency (i.e processes at a frequency of ua)- D uring

th e m ixing period, some m agnetization transfer can occur to neighbouring nuclei

such th a t some Ha which processed a t ua during ti processes at say, z/g during t2-

th e proto n Ha did not undergo any m agnetization transfer its signal will appear

at frequency coordinates of This can also occur for the nuclei, which

will show signals a t These are called diagonal peaks since they lie on the

diagonal of th e square-shaped spectrum . However, if m agnetization transfer did

occur and processed a t f/g during tg, this will generate a signal on th e spectrum

at T his is a ’cross-peak’, it is an off-diagonal signal, the m agnetization

processional frequency changed after th e mixing period (see figure 2.15).

These are th e beisic principles common to all 2D NM R experim ents. A 3D

experim ent can be envisaged as the com bination of two 2D NM R experim ents,

excluding th e detection period of th e first experim ent and the p reparation period of the second experim ent (figure 2.5).

prep. m ix.

! !

tl acquisition

J _{real tim e}

increm ental

Figure 2.5: How 2-dim ensional experim ents are combined to create a 3-dim ensional pulse sequence. Taken from A D ictionary of Concepts in NM R by S.W . Homans. 1992.

T h e tim e-dom ain d a ta are acquired during ta and a set of FIDs collected w ith

independent increm entation of b oth ti and t2-

Since th ere are three tim e periods in a 3D experim ent, it requires a 3D Fourier transform to convert all the tim e-dom ain d a ta into the more fam iliar frequency spectra.

T h e best way of representing th e d a ta is as a cube which comprises a series of 2D cross-sections of th e 3D spectrum (figure 2.6).

1

F3 = x

2.2

T he B asic Techniques

T h e following discussion a tte m p ts to explain how certain pulse sequences work, firstly for the sequences which can be regarded as building blocks and then in section 2.3 for th e m ore complex sequences which were used in this project. These explanations make use of a vector model whenever possible and are not intended to be comprehensive. In th e interest of sim plicity I will only m ake references to the m agnetization which give rise to the cross-peaks expected to be seen in the spectra; o th er m agnetization com ponents will be ignored. This also m eans I will not be discussing phase cycling.

2.2.1 T he T O C SY experim ent

T h e TO C SY (Total C orrelation Spectroscopy) experim ent is a 2D experim ent used

for shift correlation. It identifies which protons or spins are p a rt of the same