Molecular structure by two-dimensional NMR spectroscopy

Molecular structure by two-dimensional NMR spectroscopy

Journal of Molecular Structure, 173 (1988)17-30 Elsevier Science Publishers B.V.,Amsterdam -Printedin The Netherlands 17 MOLECULAR SMtLICTUREBY TWO-...

621KB Sizes 1 Downloads 167 Views

Journal of Molecular Structure, 173 (1988)17-30 Elsevier Science Publishers B.V.,Amsterdam -Printedin The Netherlands

17

MOLECULAR SMtLICTUREBY TWO-DIMENSIONALNMRSPECIR~PY

R. FREEMAN Department of Physical Chemistry, Cambridge University, Lensfield Road, Cambridge CB2 lEP, England. SUMMARY

Two examples are presented of the use of two-dimensionalNMR spectroscopy to solve molecular structure problems. The first is called correlation spectroscopy (COSY) and it allows us to disentangle a complex network of spin-spin couplings. By dispersing the NMR information in two frequency dimensions, it facilitates the analysis of very complex spectra of organic and biochemical molecules, normally too crowded to be tractable. The second application exploits the special properties of multiple-quantum coherence to explore the molecular framework one C-C linkage at a time. The natural product panamine is used as a test example: with some supplementary evidence, the structure of this six-ringed heterocyclic molecule is elucidated from the double-quantum filtered two-dimensionalspectrum.

INTRODUCl'ION It often happens that an NMR spectrum contains too much information for the purposes of structure determination and then special techniques must be found to simplify the spectrum and separate out the useful data. This is particularly true of large organic molecules and most molecules of biochemical interest. Fortunately there are many useful tricks we can play on the nuclear spins so that they yield their secrets in a simple form. This is what distinguishes NMR from most other forms of spectroscopy; it is seldom sufficient to take away a spectrum and make no further demands on the spectrometer: more usually we have to devise a whole range of supplementary experiments to make a complete assignment. Double resonance is one such technique. By irradiating the sample with a second radiofrequency field we can often identify nuclei that are spin-coupled, even though the network of coupled spins may be very complex indeed. These decoupling experiments were usually carried out by the operator setting the frequency of the second radiofrequencysource at some suitable point in the spectrum. but in modern spectometers. it

is perfectly feasible to

conduct a systematic search throughout the entire spectrum under computer control. A second powerful double resonance technique is provided by the nuclear Overhauser effect - the enhancement of the intensity of a given resonance line when a close neighbour in the molecule has its resonance response saturated.

0022-2860/88/$03.50 0 1988ElsevierSciencePublishersB.V.

18 The magnitude

of

this

the appropriate structural acids

questions

or folding

exchange,

Now it

turns

(ref.

introduction

nuclear

conditions

during

-

The process

are

signal

but it

is

NMR experiment

is performed

written

F, may be thought of

experiment

where the perturbing

that certain

the entire

two-dimensional double

involves

steps

conditions

manner from the which normally receiver

on during

t,.

inactive

The raw S(t,,t,).

transformation

In Jeener’s

respect

as a two-dimensional

follows is

matrix,

a Fourier

spectrum. with

by

the

in small

the physical

(te)

to both

ti

and t,.

transform

W-1 .F,)

-

dimensions.

in terms of an equivalent radiofrequency

The “new” double

has been varied

range of NMR responses.

resonance in a stepwise

Indeed we shall

see below

spectra

are virtually

indistinguishable

resonance

experiments.

However two-dimensional

spectroscopy

has the advantage

are

at each stage

gathered

It

incremented

we would perform

is an NMR spectrum in two frequency

computer-controlled

2).

period

switched

resonance

in two-dimensional

The spectrometer

a frequency-domain

transformation

may be formally

to cover

of double

idea was proposed

(ref.

acquisition

period.

on NMR

exchange.

categories

which is this

of atoms

the same

effect

thus in the form of a two-dimensional

dimension

fashion

by their

revolutionary

(ti)

for

by the second radiofrequency

by Ernst

During

S(t*.t,) The result

for

of amino

of some atom or group

changed in some significant

to give

Fourier

This

period

period,

data are

S(F,)

experiment

power of useful

evidence

changes within

each have analogues

generalized

during

Now in a conventional S(t,)

extremely

direct

by chemical

three broad

in 1971.

the evolution

the evolution

experimental

site

range of values.

after

transfer

of populations

these

of an evolution

precession

provides

are monitored

in the 1960’s,

introduced

prevailing

immediately

to the sixth

such as sequencing

or conformational

to another

1) and later

over a suitable

resonance

the physical

out that

developed

spectroscopy,

proportional

and it has proved

molecules,

spin populations

transferred

experiment,

for

either

and a perturbation

gets

Jeener

double

molecules,

Nuclear

intensities, field

inversely

in biochemical

fashion

between different molecule.

is

distance,

of proteins.

In a similar chemical

effect

internuclear

of

of higher

sensitivity,

since

from

strong

NMR signals

the experiment.

CDRRELATION SPECIROSCOPY Before information,

a high

resolution

NMR spectrum can be used

there has to be an assignment

of

to derive

the resonance

lines.

structural This

may

19

be based on chemical shift information or relaxation times or on the form of the multiplet fine structure. If there is a resonance A which already has a firm assignment, then it is often possible to assign another resonance X on the basis of a. spin-spin interaction JAx. The evidence might be a double resonance "decoupling" experiment where the fine structure on A is modified by strong irradiation of the X resonance. Alternatively. coherence may be transferred from A to X and from X to A in a two-dimensionalexperiment. During the evolution period. nuclei precess at frequencies fA and fX Hz. During the acquisition period, these frequencies persist, but there is also The two-dimensional some coherence transfer f -fXand fX.f A' A spectrum thus contains two types of response, signals that lie on the principal diagonal (Fig. 1) where F, = Fe, and off-diagonal signals or "cross-peaks" arising from coherence transfer.

Fig. 1 Schematic correlation spectrum of a system of two protons A and X.

The positions of these cross-peaks indicate which resonances are related by spin-spin coupling because as JAx -0

the cross-peaks disappear. This

correlation spectroscopy (CD%) is a powerful method of assignment (ref. 3). It may be used in homonuclear systems, relating different protons in the same molecule, or in heteronuclear systems, relating (say) carbon-13 resonance to proton resonances.

A Practical Illustration Proton-proton correlation spectroscopymay be illustrated by reference to the hydroxytricyclodecanedionederivative shown in Fig. 2, which contains eleven protons exhibiting a complex network of couplings.

The

20 two-dimensional is

spectrum is also

the principal

diagonal

one-dimensional of

cross-peaks

is

projection

proton-proton

facilitated

the COSY spectrum: was obtained

There is a wealth

spectrum.

the 28 observable

it

onto one of

The first

2.

of cross-peaks,

chemical

the frequency

shifts

of

followed

H

45

proton

Some of they are

the proton-proton

the envelopes

of

couplings

H

H “t H

be observed

in this

network derived

manner are

H

inset

molecule

OH

A

on the right.

are

small

because

the instrumental

from these

signals

“.H

0

the top margin.

in this

of the CXBY experiment

the two time-domain

to light

the molecule

along

and they may be hidden within

may nevertheless

the timing parameters

were brought

COSY spectrum of

spectrum is displayed

long-range,

Cross-peaks

coupling

. .. H H

H/’

Two-dimensional

by

axes.

I

2.

This

but no J-splittings.

of the CDSY experiment

H

decoupled

some

the various

the top margin of

0

Fig.

to notice

reflecting

Identification

couplings.

modification

thing

the conventional

by the spectrum which runs across

has proton

by a slight

shown in Fig.

(where Fl = F2) which carries

tiny

couplings

linewidth. by altering

(ref.

3) and by suitably

S(t1)

and S(t,).

shown dotted

from the two-dimensional

shaping

Coup1 ings

in the schematic

CCSY spectra

(Fig.

3).

that

21 COSY spectra

have been extensively

Often

spectra.

the detection

identifiable

protons

once and for

all.

Fig.

Schematic

3.

Solid

lines

will

be enough to solve

diagram

indicate

used to disentangle

of an appreciable

of

the couplings

resolvable

complicated

coupling

between

a particular

chemical

in the molecule

couplings;

dashed lines

proton

two problem

shown in Fig.

long-range

2.

couplings.

Pseudo-CDSY One disadvantage data matrix otherwise

finer

is

sampling

on protons

digital

ruled

indeed,

out because

it

in order

involving

t,

one

structure In practice

increments

and an

unrealistically

resonance

experiment

long experiment

a homonuclear

to examine the fine

would make the number of

a very

the final’

to have appreciably

caused by overlap.

ambiguities

that

Typically

is violated.

4kHz x 4kHz in 4Hz steps.

than this

or to resolve

is

NMR spectrum in both dimensions

Sometimes we would like

storage.

resolution

entailing

condition

might cover

words of data

the cross-peaks

COSY experiment

must encompass the entire

the Nyquist

experiment million

of the two-dimensional

S(F,,Fe)

very

on this

large

large

data

matrix. This

is where the analogous

advantage.

Since

than by Fourier

transformation

to examine a narrow is

involved.

excite

only

is already across

double

the Fi domain is explored

region

A selective one resonance known from prior

the chemical

shift

of a time-domain

in very

small

radiofrequency line

by incrementing

pulse

is used

is

quite

sampling

(ref.4)

pulse

and the rest

of

rather

feasible condition

calculated

the conventional

the selective

range of interest

it

no Nyquist

Since

at a time.

experiments,

signal,

steps;

can be used to a frequency

to

NMR spectrum

can be stepped the F, domain can

22

be ignored. This has the effect of a "zoom" operation, where a chosen cross-peak can be examined with very fine digital resolution, without having to perform an impossibly large number of experiments. We call this experiment pseudo-CDSY in order to distinguish it from a true two-dimensionalFourier transform experiment (ref. 5).

Fig. 4 shows one of the cross-peaks of the

pseudo-CQSY spectrum of a three-spin system (2.3-dibromopropanoicacid) examined under high digital resolution.

:::: ::::

/

::: :::

00

00 /

...** ... . .

. / .,I

. .

,

,

G .I

I

.

.

//

‘::: .:::

. . ::: . .

. .

:::: ::::

30 Hi

\

a 0

b

0” l

0

00

00

/

20 Hz 475 Hz x 475 Hz Fig. 4. Zoom operation on a cross-peak in a pseudo-CDSY spectrum.

MULTIPLE-QUANTUM COHERENCE

The selection rule for NMR spectroscopy is that the magnetic quantum number should change by only one unit, Am = k 1. Multiple-quantum transitions are formally forbidden. When we examine this restriction in more detail, we find that with pulse excitation schemes the selection rule can be violated in as much as a coherence can be excited between states differing in quantum number by more (or less) than one unit, provided more than one excitation pulse is used. The catch is that these multiple-quantumcoherences induce no detectable signal in the spectrometer. Whereas normal signals can be thought of as arising from the alternating voltage induced in a coil by a rotating magnetic dipole, multiple-quantumcoherences behave more like quadrupoles and induce no signal at all.

23

-Two-dimensionalspectroscopy provides the key to this problem. The evolution period ten now be put to good use. We can excite multiple-quantum coherence and allow it to precess during the evolution period: it may be reconverted to observable nuclear magnetization detected in the normal manner. The multiple-quantum frequency is never observed directly but is deduced by its indirect effect on the observed NMR signal. It is "sampled" by varying the evolution period (when the spectrometer receiver is inactive).

CARBON-CARBON CONNEClIVI'lY

There is one application of double-quantum coherence which provides a powerful tool for molecular structure determination. If the carbon framework of a molecule can be established, the molecular structure is virtually determined because linkages through heteroatoms (nitrogen or oxygen) are relatively infrequent and can usually be inferred from other evidence. Now it turns out to be quite fortunate that Nature provides carbon-13 in low abundance (1%). In one molecule in lo* we would therefore expect to find adjacent carbon-13 nuclei. Directly-bonded carbon-13 atoms show a spin-spin coupling of 30-40 Hz whereas all longer-rangecouplings are much smaller and can be neglected. If we can detect these carbon-carbon couplings we can therefore establish connectiviteand thus follow the carbon chain step-by-step, identifying chain branching and ring formation. We may therefore trace out the entire carbon framework of the molecule, reaching a 'dead-end' only when the chain terminates or connects to a heteroatom such as nitrogen, oxygen or sulphur. The evidence is direct and unequivocal; if a carbon-carbon coupling of appreciable magnitude is detected, the two carbon atoms must be immediately adjacent in the molecule. If a given carbon atom shows three such couplings, this indicates chain branching. If a given carbon atom gives evidence of a chain of n such linkages which loop back to the original site, then an n-membered ring is present. The important point is that each piece of evidence for carbon-carbon coupling is independent of all others; molecules with three or more carbon-13 spins are of such low abundance that they can be safely neglected. The principal disadvantage of the method is the poor inherent sensitivity since only one molecule in lo* contains the useful information. Fortunately, technical improvements in NMR probes and magnets are steadily increasing the intrinsic sensitivity of the spectrometer.

The great advantage of the

two-dimensionaldouble-quantum experiment is that the connectivity information is unequivocal.

24 Exnerimental Method The first step is to extract the useful carbon-carbon coupling information from under a much stronger carbon-13 signal from molecules containing a single carbon-13 spin. This is where double-quantum coherence is useful. Two coupled spins can generate double-quantum coherence but isolated spins cannot. Since a radiofrequencyphase shift affects double-quantum coherence twice as much as single-quantumcoherence, it is possible to set up a phase-cycling scheme which suppresses the unwanted signal components very effectively leaving behind the signals that have passed through double-quantum coherence. This leaves a four-line AX or AB-type subspectrum for each pair of directly-bonded carbon atoms. These subspectra are then separated in the F, dimension according to the frequency of the double-quantum coherence which evolves during t, (ref.6). This double-quantum frequency is given by (6A + 6X) Hz and is essentially unique to each pair of coupled carbon-13 spins. We can therefore simulate a typical two-dimensionalspectrum viewed as a contour map (Fig. 5).

Fig. 5

A typical subspectrum from a two-dimensionaldouble-quantum

experiment.

There is a four-line spectrum running horizontally, showing chemical shift 6A and 6X and splittings JAx, with its centre-of-gravityat (LjA+ bX)/2. The Fi ordinate is the double-quantum frequency (hA + 6X). All other subspectra must have their centres-of-gravityon the diagonal line of slope 2.

These

restrictions impose a characteristicpattern on the two dimensional spectra, allowing connectivities to be deduced even the

presence of a great deal of

noise.

The Structure of Panamine The method is illustrated by consideration of a real case of structure determination, the natural product panamine, one of the Ormosia alkaloids

(ref.

7).

The conventional

a through start

t in order

tracing

resonance of

of

out the carbon

with a very

reducing

at a rate

which allows

too distant

for

this

relaxation

d

II

1

!

to site

ef

The conventional dimension). lines;

four

chemical

represents

(indicated

of

carbon-13

of

the carbon-13

are

on the diagonal

by the dotted

region

hi

providing

J’klmn

04

double-quantum

centres

the diagram

i.

It

is

therefore

carbon d, carbon

(the

F,

by horizontal of gravity

The ordinates

in the Fi dimension.

to carbon a, lines).

the top of

their

dashed line.

to site

spectrum of panamine.

have been highlighted

shown, with

(6A + ox)

i -' i'

41

spectrum runs along

frequencies

attached

This

protons

spectrum of panamine is

the relevant

k y' 0'i 'i / 0/ /' ,;,' i,

subspectra

one connectivity

atom is directly

the nearest

relaxation.

,

such subspectra

shifts)

repeated

between pulses.

0 ,. , , ,

The four-line

double-quan&m

is

1

: : i4

of part

has the effect

excitation

with

Q

;

A tracing

to

i.

i',

Fig.6

site

double-quantum

6 shows a tracing

This

time.

time for

spin-lattice

labelled place

i which has a carbon-13

the pulse

i is a quaternary

the connectivity

c

site

since

two-dimensional

20 resonances A convenient

shielding.

relaxation

line

efficient

but Fig.

ab

5

of

insufficient

The experimental

evidence

framework is

that carbon

to provide

complex,

spectrum contains

magnetic

long spin-lattice

the intensity

is good evidence

quite

carbon-13

increasing

(mean

represent

the

Each subspectrum clear

that

this

e and carbon n

carbon

26 We can therefore site

1 connected

Fig.

7.

First

stage

The next step chains

directly

is

8

site

(Fig.

the evidence

9).

One reason

of

Next it

spectrum to see if

these

a end site

m,

is noted that both m and j are

is

incomplete

for

atoms.

connectivities

finding

indicating

the presence

continues,

but since

the spectrum.

Now since

In fact

sites

and draw in a “partial”

write

“partial”

since

and four

unconnected

so many “dead ends” is

of carbon atoms attached

somewhere else.

7).

on penamine.

found between site

carbon atom h.

of following

case,

region

are

framework showing

end n (Fig.

of a

of a six membered ring.

(Fig.

resonances

a.d.e

8).

real

nitrogen

j.

to the m

Formation

contains

the molecular

sites

determination

Connectivities

n and site

ring

The process

of

to examine the two-dimensional

attached

six-membered

to draw part

of connectivity

ten be extended.

and between

Fig.

begin

to the four neighbouring

nitrogen

we can assign

nitrogen

segments are

is a found

that panamine

is electron-withdrawing,

to nitrogen

the same nitrogen

this

atom at

all

fall

in the low-field

resonances

a through g to such

the end of

these

atom may be terminating

chains.

We

another

chain

27

f 8P

Fig. 9

Four apparently unconnected segments of the panamine framework.

A second reason why a connectivity might be missed arises out of a shortcoming of the double-quantum experiment itself. The pulse sequence has to be set with a timing parameter chosen so as to optimize the creation of double-quantum coherence, based on the known range of J,, values. This adjustment occassionally proves to be incorrect if the two carbon chemical shifts are very similar, for the spectrum is then a strongly-coupledAB system rather than a weakly-coupled AX system. The corresponding subspectrum is of low intensity and may not be observed. Fortunately the strongly-coupledcase has been analysed (ref. 8) and it is known that the intensity of the AB subspectrum can be restored by repeating the experiment with a longer setting for the timing parameter. In the panamine example, carbons 1 and m form such a strongly coupled AB system, as do carbon atoms p and q.

These

connectivities are shown dotted on the next version of the molecular framework (Fig. 10).

From elemental analysis it is known that panamine contains three

heterocyclic nitrogens, so it is now possible to tie together the loose ends of the previous diagram. although of course this is no longer a unique

28 solution to the structural problem. Supplementary NMR evidence can now be used to check whether all the carbon valencies have been satisfied.

Fig. 10. Result of assigning attachments to nitrogen at carbon sites a through g on the basis of their low-field shifts.

It is relatively straightforward to determine the multiplicities of each carbon site, indicating the number of directly attached protons, so it is known how many C-C or C-N bonds to expect at each site. Consequently it turns out that site b is short of an attachment and it has to be concluded that this carbon atom is attached not to one nitrogen but to two. With this information it is possible to draw up possible molecular frameworks for panamine. one of which may be preferred because it contains only six-membered heterocyclic rings (Fig. 11).

One of the nitrogens must of

course be an NH group because it has only two C-N bonds. Although this shows the proposed connectivity, it is not the normal way that the structure would be presented.

The organic chemist might rearrange this drawing to emphasize

29 the cyclohexane-type

Fig.

11. Final

Fig.12

rings

as illustrated

framework proposed

The structure

for

in Fig.

12

panamine

of panamine as an organic

chemist might draw it.

30 REFERENCES 1.

J. Jeener, 1971.

Ampere International

2.

W. P. Aue,

E. Rartholdi

3.

A. Bax and R. Freeman,

4.

C. Rauer. R. Freeman, Reson. 58 (1984) 442.

5.

S. Davies.

6.

A. Rax, R. Freeman, ( 1981) 478-483.

7.

T. H. Mareci

8.

A. E&LXand R. Freeman,

J.

Friedrich T.

Summer School,

and R. R. Ernst, J. Mag,n. Reson. T. A. Frenkiel,

and R. Freeman,

J. Chem. Phys. 44 (1981)

J. Keeler

and R. Freeman, A. Frenkiel

Rasko Polje.

and A.

and H. H. Levitt,

J. Magn. Reson.

48 (1982)

41 (1980)

64 (1976)

2229.

542-561.

J. Magn. Reson.

J. Magn. Reson.

Yugoslavia,

597.

J.

Shaka,

(in

J. Kagn

press).

J. Magn. Reson.

158-163.

43