MRRC Structure Elucidation Notes

    Contents

• Important notes
  
  
  • Total recycle delay, DR = D1 + AQ
    
    
  • Signal-to-noise (S/N) considerations
    
    
  • Using NUS or LP
    
    
• Basic experimental sequence
  
  
  • Proton 1D
    
    
  • Carbon 1D
    
    
  • Inversion recovery to determine average T1
    
    
  • 1D and 2D TOCSY
    
    
  • COSY
    
    
  • HSQC
    
    
  • HMBC
    
    
  • H2BC and 1,1-ADEQUATE
    
    
  • LR-HSQMBC
    
    
  • NOESY and ROESY
    
    
• Analysis of carbon skeleton
  
  
• Determining stereochemistry
  
  
• References

Back to Contents

  Important notes

• Bruker parameter sets use total number of points in F1 and F2 dimensions while the MNova and Burns paper both use number of complex points which is 1/2 of the total points.
  
  
• Preferable probe: optimized for 1H detection
  
  
• Preferred tubes:
  • Use 5 mm tubes in 5 mm probes.
  • Use 3 mm tube only if sample contains high salt
  • Use Shigemi tubes - for mass-limited samples
    
    

Back to Contents

---

 

Total recycle delay, DR = D1 + AQ

Check your pulse sequences that they do not use purge pulses in the beginning of D1. If they do, you cannot utilize T1 relaxation that occurred during AQ time. This is most important for HMBC because AQ is long.

• DR = 1.26 T1(average)
  
• Use inversion recovery pulse sequence to determine average proton T1
• T1 reduces with Mw in a small molecule range (then begins increasing)
• However, at high field, the point when it increases shifts to lower Mw
  
  
• A rule of thumb at 400-500 MHz *):
  
  

Mw, Da	If Mw is unknown, use # protons **)	Recycle delay, DR, seconds
< 275 Da	< 30 protons	2 sec
275-350 Da	30-40 protons	1.5 sec
> 350 Da	> 40 protons	1 sec


*) This table must be modified for use at ultra-high fields because the T1 minimum shifts to smaller Mw.
**) If Mw is unknown, estimate number of protons from 1H 1D integration


• 
  

 

Back to Contents

---

 

Signal-to-noise (S/N) considerations

• Sample concentration and number of transients
  
  
  • If you change of NS by a factor of N, the S/N changes by a factor of sqrt(N).
    
    
  • Always know and record your sample concentration in the experiment (in mM). It helps understand how much signal you expect and relate to other experiments. Simple rules of thumb are:
    
    
    • If you double your concentration, your S/N doubles. If you were happy with S/N at the original, lower concentration, you may reduce NS by a factor of four.
      
      
    • If you drop your concentration by a factor of two, to keep your S/N as in previous experiment, you must increase NS by a factor of four.
      
      
• 2D experiment: a compromise between the number of scans, NS, versus number of increments, TD F1, if one wants to keep the same total experiment time
  
  
  • The correct way of thinking about S/N in a 2D spectrum is to consider S/N per one increment. That is the number for an overall S/N divided by TD F1.
    
    
    • Example 1: If you want to keep the same acquisition time but increase resolution in F1 by a factor of two (double TD F1), you will have to reduce NS by a factor of two. As a result, S/N per increment will reduce by a factor of sqrt(2).
      
      
    • Example 2: if you reduce resolution in F1 by a factor of 2, you may increase in NS by a factor of 2 to keep the same experiment time. In this case, the S/N in the spectrum will increase by sqrt(2).

 

Back to Contents

---

 

Using NUS or LP

• MISCONCEPTION WARNING:
  A widespread perception (including in [1] and [2]) is that NUS gives extra resolution at no cost. This is not true.
  
  
  • If you acquire 1/N fraction of the full plane and use NUS to reconstruct to the full experiment, the S/N in the reconstructed full plane will be smaller by a factor of sqrt(N) relatively to the traditional experiment that acquired the full plane.
    Example: 25% NUS results in two-fold reduction in S/N (because we increase TD F1 by a factor of four during reconstruction)
    
    
• A strict requirement according to Delaglio [2]: "The number of NUS increments (actually acquired) must be about ten times greater than number of cross-peaks (TD F1 > 10x N(Peaks)) for accurate representation of spectra".
  
  
  • NOTE: This means distinct peaks in the F1 projection, or distinct chemical shifts in the indirect dimension.
    
    
  • Workflow
    
    
    • Count the number of distinct chem shifts along F1 (= X-nucleus resonances) and multiply by 10. This is a minimum number of TD F1 elements you have to record whether as a traditional plane or NUS.
      
      
    • If this TD F1 is less than the resolution you want, you will either
      • increase TD F1 to a desired resolution and acquire traditional planes
        or
      • switch to NUS and adjust percentage (or fraction 1/N)
        IMPORTANT: S/N will drop as sqrt(N) (see below)
        
        
  • IMPORTANT: Broad peaks must be counted as several peaks and will require more elements to represent them!
    
    
• To control sensitivity of the experiment while using NUS, you must answer this question:
  
  
  • Can I afford to lose a factor of two in S/N yet retain enough sensitivity to see all peaks in my 2D spectrum?
    If answer is yes, you may proceed with NUS at 25% (see requirement of a minimum TD F1 above!).
    Example: small molecules at very high concentrations (>100  mM).
    
    
  • In general, if you can tolerate a reduction in S/N by a factor of N, you can use 1/N^2 fraction of a full plane with NUS acquisition (subject to a requirement of a minimum TD F1 above!).
    
    
  • IMPORTANT CAVEAT: Both NUS and LP will favor larger peaks while degrading the small ones.
    
    
  • NOESY and ROESY are thus much more challenging: their most important peaks are small. These experiments are better be acquired as full traditional planes for quantitative assessment or with NUS no less than 50%.
    
    
• Advice from Delaglio:
  • (with NMRPipe SMILE) you may use NUS to extend traditional experiment instead of using LP. It gives fewer artifacts (but may not work in MNova).
    
    

To do

• Create a workflow to use NUS in NMRPipe to have more flexibility in analysis

 

 

Back to Contents

    Basic experimental sequence

• NOTES:
  
  
  • Atoms connected to 14N may be broadened in some cases due to quadrupolar relaxation.
    
    
  • Peaks broadened by exchange are better observed with a large (up to 50 Hz) EM broadening .
    

   Proton 1D

• Use two kinds of experiments:
  
  
  1. Quick experiment
     1. short DR (D1=0.1, AQ=2) for monitoring sample and spectrometer stability. Record a large spectral width to ensure there are no "surprise" peaks (and they don't appear over time).
     2. When a previous experimental series is available, overlay the experiment obtained with the same NS/gain/D1/AQ to estimate whether the sample concentration is similar or not and make necessary adjustments in acquisition time of all experiments.
        
        
  2. A fully relaxed experiment with a total relaxation delay, DR, set to 10 sec (D1=6, AQ=4).
     
     1. This experiment will be used for integration.
     2. Spectral width may be adjusted to only cover the peaks of interest +1 ppm from each side.
     3. Increase NS as needed to have better S/N for integration.
        
        
• NOTE: Cryoprobe is difficult to shim well so the proton peaks may be showing distortions/splittings on the order of 2 Hz. Set 1Hz broadening in apodization to mask the distortions.
  
  

   

  Back to Contents

  ---

    

  
  
  

    Carbon 1D

From the book:

• Use DR= 10 sec if MNova Verify will be used
• Set DR=5-7 sec and use zg30 sequence for bringing up unprotonated carbons
• Generic carbon experiment is run with zg30  and DR of 2.5 sec (D1=2sec).
  NOTE: If you want to integrate you will have to remove NOE from carbon pulse sequence!!!!

NOTE: Do not skip it: acquiring carbon is important because it allows to detect if the sample is a mixture of stereoisomers!

NOTE 2: This experiment also shows whether you can cut spectral width down to cover a narrow crowded range. If you don't have downfield shifted carbons - cut SW in HSQC, HMBC, and H2BC to improve carbon resolution!

NOTE 3: My practical experience: full 90 and shorter recycle do more for slowly relaxing carbons than long wait during the same total experiment time. Try these parameters:

Back to Contents

    Inversion recovery to determine average T1

• If amount of compound is limited and acquisition will be lengthy, it is better to determine the experimental T1 to tune recycle delay better.
  
  
• Simple inversion-recovery experiment:
  • series of some 5 points is enough
  • Look for the same number of positive and negative peaks: it gives t(null)
  • Estimate T1 from that trace :  T1 = 1.44 * t(null)
    
    

Setup notes

• Edited chapter from Topspin manual:  Protocols_and_Manuals/Bruker_manuals/topspin/topspin_advanced_nmr_experiments_chapters/t1.pdf
  
  
• Make sure to:

• issue 'pulsecal' in T1 experiment to set 90-degree pulse
• if you change number of T1 values, enter the same value TD F1 field
• phase first fid upside down

 

Back to Contents

---

 

 

  1D and 2D TOCSY

• If your sample is a mixture of multiple compounds, TOCSY helps establishing peaks belonging to isolated spin systems.
• If your sample is a well-defined result from purification, TOCSY is not necessary, you can use COSY instead to mark off peaks of impurities
  
  
• To resolve overlapped multiplet structures
  • use 1D TOCSY with selective excitation
  • must use a zero-quantum filtered experiment for undistorted multiplets
    
    
• Quick proximity measurements with selective 1D TOCSY
  • five spectra with mixing time ranging from 10 to 80 ms will allow proximity estimates (along the carbon skeleton).
    
    
• Use a mixing time of 60-80 ms for a routine experiment.
  
  

TOCSY setup notes

mlevphpr.2 - 2D TOCSY with solvent presaturation - works well

PROCESSING NOTE: When processing first 1D, amide peaks on the left are always negative yet positive in 2D. Not sure why but this is normal.

 

mlevgpph19 - 2D TOCSY with solvent suppression using a binomial 3-9-19 sequence - works well

 

PROCESSING NOTE: When processing first 1D, amide peaks on the left are always negative yet positive in 2D. Not sure why but this is normal.

 

Back to Contents

    COSY
 

• COSY-45 has advantages over COSY-90 (despite being less sensitive)
  • diagonal is narrower - helps resolving peaks near a diagonal
  • cross-peaks are tilted. Useful observation is this:
    • cross-peaks between methylene protons (most of the time) for
      • geminal protons (two-bond) have positive tilt
      • vicinal protons (three-bond) have negative tilt
        
        

• DQF COSY removes singlets (and their t1 ridges) and allows estimating coupling constants
  • active splitting in a cross peak shows as antiphase splitting <<< may be measured through a cross-section through the peak
  • all other splittings - in-phase (up-up or down-down)
    
  • Less sensitive than COSY-45
    
    
• Long-range COSY peaks may be observed for these extened planar structures:
  
  
  • 
    

Back to Contents

    HSQC

• Edited HSQC has a risk of incorrectly phased or nulled peaks if 1JCH is different from a (standard) 145 Hz. There is a version tolerant to variation in 1JCH [6].

• CAUTION: This experiment uses carbon decoupling.
  • Do not use AQ longer than 0.2s to avoid heating of coils and probe damage
  • On a CryoProbe watch NMR Coil Heater Power:
    • it may drop to as low as 18% but should not be lower'
    • It must be stable over time;

• HSQC Spectrum is mostly diagonal, may be easily folded to increase resolution. However, it will require manual un-aliasing the folded peaks which is not easy for beginning user.

Back to Contents

   HMBC
 

• The least sensitive of 2D experiments. Use about 2x NS than HSQC
  
  
• Use longer AQ because signal in HMBC appears as an echo, in the middle of a fid and there is no decoupling
  • 0.3 sec for Mw <= 400 Da
  • 0.4 sec for Mw > 400 Da
    
    
• Multiple-bond J(CH) setting
  • common:  8 Hz. Good peaks for couplings with 3-4 Hz but will miss smaller couplings
  • optional:  4 Hz will detect more peaks for small couplings but may miss stronger ones. Also, - less sensitive.
  • CIGAR [4] and IMPEACH [5] accommodate broad range of coupling constants but have lower sensitivity in general
  • CIGAR samples coupling constants  from 4 to 10 Hz:
    • works better for small molecules < 350 Da
    • Use HMBC for >350 Da
      
      
• For nitrogen-containing compounds
  • 13C-HMBC can connect to NH and NH2 (peptide N-terminus) but only if solvent is not D2O
  • 1H-15N HMBC is essential to connect surrounding protons to nitrogens
    
    • This experiment needs a lot longer acquisition time (100x?) than 13C-HMBC for the same sensitivity
    • CIGAR sequence is best choice because it can accommodate broad range of JNH constant values [4]
      
      

Back to Contents

   H2BC and 1,1-ADEQUATE
 

Both allows to find 2-bond peaks for olefin and aromatic fragments that do not show up in HMBC

H2BC

• CAUTION: This experiment uses carbon decoupling.
  • Do not use AQ longer than 0.2s to avoid heating of coils and probe damage
  • On a CryoProbe watch NMR Coil Heater Power:
    • it may drop to as low as 18% but lower it should not be
    • It must be stable over time
      
      
• NOTE: Sensitivity of H2BC is lower than HSQC. Use, at least, 2x NS and 1/2 of TD1 of HSQC for a useful dataset.
  
  
• CAUTION: Experiment uses COSY-type transfer from main proton to a proton on the neighboring carbon. Therefore, the H2BC only detects 2nd carbons if they are protonated! Unprotonated carbons do not give cross-peaks.
  For example, H3C-C(quarternary) will not yield a cross peak for H(methyl)- C(quartenary). Also, couplings with small J couplings will be missing.
  
  
  

1,1-ADEQUATE

• Unambiguous determination of 2-bond cross peaks for both protonated and non-protonated carbons.
  
  
• 100x less sensitive than HMBC because uses 13C-13C pairs
  
  
• Optimization
  • 1JCH = 145 Hz- common choice
  • CC-transfer. Coupling constants range from 30 to 80 Hz (aliphatic - 30-45, aromatic 55-70, olefinic - up to 80 Hz).
    Use constants:
    • JCC=60 Hz for compounds with both aromatic and aliphatic
    • JCC=40-50 Hz for compounds with no aromatic or olefino groups
      
      
• Be careful with S/N optimization. Using TD F1 too small may defeat the purpose - resolution in F1 will be too bad.
  
  

   

  Back to Contents

  ---

   

   

  
  

   LR-HSQMBC
 

• most useful for proton-deficient molecules (H/X <1)
• optimized for 2 Hz or 4 Hz couplings
• 2Hz gives up to six-bond correlations
• must be used to supplement HMBC because distinction between 2-bond and longer cross-peaks is not clear
  
  

Back to Contents

      NOESY and ROESY
 

• D1 optimization by using AQ time will not work here: most experiments will have a purge pulse in front of D1! Therefore, relaxation during AQ cannot be used. Use D1= Recycle time needed by this molecular weight.
  
  
• Utility:
  • NOESY up to 400 Da
  • ROESY (better choice) for 400-600 Da but NOESY may still be used
  • ROESY > 600 Da
  • NOESY again at much higher Mw (several kDa), more viscous solvents, higher fields.
  • IMPORTANT: At higher magnetic field the NOESY zero point shifts to lower Mw! That means ROESY becomes more useful sooner.
    
    
• Cross peaks in NOESY and ROESY:
  • NOE: opposite sign as diagonal peaks for Mw below the null point, the same sign as diagonal for large Mw
  • ROE : always opposite sign as diagonal regardles of Mw
  • EXSY, exchange: the same sign as diagonal.
    • Problem: EXSY peaks may overlap with cross-peaks cancelling each other.
    • Exchange with water - cross-peaks at water frequency
  • COSY artifacts: split cross peaks with positive and negative components.
    • Helpful trick: Using maximum mixing time maximizes NOE/ROE but not the COSY, therefore the COSY artifacts may be identified.
    • 1D selective NOESY and ROESY: Integration cancels out COSY artifacts for well-resolved cross peaks.
  • TOCSY artifact in ROESY: use transverse-ROESY experiment to remove it.
    IMPORTANT: The header of a tr-ROESY sequence in Bruker library incorrectly states that it is CW experiment (regular ROESY). This is a copy-paste glitch, look for a ROESY part in a pulse sequence: you will see a pulse train in a loop for tr-ROESY and plain low power proton pulse for CW ROESY.
    
    
• Mixing times in NOESY should be around average T1 of a molecule:
  • 2-3 sec for Mw < 250 Da
  • 1 sec for 250 < Mw < 400 Da
  • 0.5 sec for Mw >400 Da
    
    
• Mixing times in ROESY
  • 0.5 sec maximum on room-temp probe
  • 0.8 sec maximum on a CryoProbe (watch the NMR coil heater not to drop below 18%)
    
    
• Selective 1D NOESY/ROESY
  • specific distances may be quickly checked as it is easy to acquire a series of mixing times in a short time

NOESY setup notes

• 2D: Bruker chapter  2D_NOESY.pdf
  
  
• Selective 1D: Bruker chapter 1DselNOESY.pdf

ROESY setup notes

roesyphpr.2 - 2D transverse ROESY with solvent presaturation. - it is less sensitive than CW roesy but removes TOCSY and false-ROE

NOTE: The sequence header incorrectly states it is CW ROESY! Presence of p25 pulse with a loop indicates transverse ROESY version.   

 

 

roesygpph19 - ROESY sequence with 3-9-19: works best for water suppression! 
NOTE: The sequence is CW ROESY! Beware of TOCSY artifacts!   

roesygpph19.2 - T-ROESY sequence with 3-9-19 works but produces a spectrum with an artifact: orthogonal diagonal - cannot understant how to fix. The sequence header incorrectly states it is CW ROESY!

roesyesgpph - ROESY with selective water pulse works but produces a spectrum with an artifact: orthogonal diagonal - cannot understant how to fix

 

 

 

Back to Contents

---

 

    Analysis of carbon skeleton

• Preparation steps
  
  
  • Process proton spectra with enough line-broadening to hide shimming imperfections
  • Revealing broad peaks may need separate processing with large broadening constants 10-50 Hz
  • Make carbon and proton smooth enough (5-8 points per peak) : zero-fill double or quadruple
  • Zero-fill all 2D by, at least, a factor of two.
  • Zero-fill HSQC, H2BC, and HMBC data in F1 excessively to have smooth peak maxima easy to align
  • Process NOESY and ROESY with 1 Hz broadening to enhance sensitivity on weak cross-peaks
  • Process HMBC data to produce magnitude mode in F2 and then smoothen the splittings with a broader envelope: this allows to measure peak centers better and intensity at a peak marker will be reflective of intensity of a peak (otherwise, intensity at a peak marker is useless in a table in MNova).
    
  • Determine where (1) residual solvent, (2) major impurity peaks, and (3) 1-bond doublets in HMBC must be and mark them off.
  • Initial acquisition may include COSY and 13C-HSQC to establish molecular fragments composed of protonated heteroatoms.
    
  • IMPORTANT: Mark off peaks of impurity spin systems first to ensure they are not picked in HMBC analysis later!
    
    
    
    
• Step 1: Look at methyls HMBC first - the strongest signals in the HMBC spectra
  
  
  • Rule 1: Methyls must show all 2- and 3-bond HMBC peaks expected for the proposed structure.
    If you don't find all 2- and 3-bond HMBC, the structure proposal is strongly suspect.
    
    
  • if you see a strong methyl HMBC that is not expected from the structure: your structure proposal is certainly wrong!   
    
    
    
    
• Step 2: Collect HMBC/H2BC for CH and CH2

  NOTE: Generally, cross peaks corresponding to small J couplings will be missing.

• It is useful to look at HSQC and mark off the residual one-bond HMBC doublets not to confuse them with multi-bond HMBC peaks in analysis
  
  
• Missing HMBC peaks for CH and CH2 may not be a serious problem - often happen
  
  
• Cross-check with H2BC to correctly mark 2- and 3-bond correlations.
  CAUTION: H2BC uses COSY-type transfer from main proton to a proton on the neighboring carbon. Therefore, the H2BC only detects 2nd carbons if they are protonated! Unprotonated carbons do not give cross-peaks in H2BC.
  For example, H3C-C(quarternary) will not yield a cross peak for H(methyl)- C(quartenary).
  
  
  
  
  
• Step 3: Examine the HMBC correlation table: sort by intensity
  
  
  • All strong HMBC peaks must correspond to no longer than 3 bonds separation
    
    
  • Any strong four-bond HMBC indicates that the structural proposal is certainly wrong.





• Step 4: Re-examine all spectra to ensure that impurity peaks are not picked
  
  

• Look at TOCSY and COSY, also 1D proton and carbon for impurity peaks
  
  
• Determine where residual solvent peaks and 1-bond doublets must be and mark them off
  
  
  • Hint: Solvent doublets appear at proton shifts of C13 satellites of the solvent peak!
    
    
    
• Step 5: Add very long connectivities
  
  
  • Use LR-HSQMBC to establish very long connectivities up to six bonds in compounds with low H/X ratio (<1)

Back to Contents

   Determining stereochemistry

• Beware of strong couplings: splitting will not be equal to a coupling constant (watch for "roof pattern" of the multiplets). Full spin system simulation is needed to analyze such a coupling.
  
  
• NOE-based determination of distances
  • record NOE build-up curves with multiple mixing times
  • measure slopes of the linear regions
  • use PANIC method to correctly extend slopes to longer mixing times (see references in [1])
    
    
    
• New methods:
  
  
  • RDC  and RCSA (see [1] for references)
    • use polymers for different solvents; commercial kits are available
    • PBLG - for CDCl3, easy to make
    • look up Thomas Williamson
      
      
  • DFT calculations of chemical shifts for candidate structures
    
    
• Beware of exchange dynamics
  • in fast exchange, internuclear distance will be strongly weighted towards the conformer with a smaller distance (because of r^6 dependence).
  • in slow exchange, EXSY peaks may help to identify exchanging resonances
  • intermediate: need to use range of temperatures to get out of this regime
    
    
• Mixtures of closely related compounds
  • getting a good carbon spectrum is critical: it may allow to see doubling of peaks with similar ratios indicative of two forms. Assignment may be made by relative intensity.
  • doublets and singlets are, typically, farther away from the chiral center
  • well resolved peaks are around the chiral center
  • in this situation, HMBC will be highly ambiguous - too many protons correlated to a single carbon peak, which would be an unresolved doublet of carbon peaks from two forms.

Back to Contents

   References

1. Analysis of cross-correlations and structure elucidation; parameters of acquisition
Burns and Reynolds, "MInimizing risk of deducing woring natural product structures from NMR data", Magn Reson Chem, 2021, 59: 500-533

2. NUS requirements
Delaglio, F. SMART NMR Symposium 2021

3. Relationship of NUS, spectral resolution, and signal-to-noise ratio
Kovrigin, EL (unpublished)

4. CIGAR: HMBC sequence with tolerance to broad range of coupling constants (beware: lower sensitivity!)
C.E. Hadden, G.E. Martin, V.V. Krishnamurthy, Magn Reson Chem 2000, 38, 143

5. IMPEACH: HMBC sequence with tolerance to broad range of coupling constants (beware: lower sensitivity!)
C.E. Hadden, G.E. Martin, V.V. Krishnamurthy, Magn Reson Chem 1999, 140, 274

6. Multiplicity-edited HSQC version tolerant to variation in 1JCH [6]:
H. Hu, K. Krishnamurthy, Magn Reson. Chem 2008, 46, 683

7. The most comprehensive book:
Burns DC, Reynolds WF, Optimizing NMR Methods for Structure Elucidation, Series: New Developments in NMR vol. 17, Royal Society of Chemistry, 2019

The general (and very useful) textbook:
High-Resolution NMR Techniques in Organic Chemistry, 3rd Edition, by Timothy D.W. Claridge. Elsevier Science (May 27, 2016), ISBN-10 ‏ : ‎ 0080999867, ISBN-13 ‏ : ‎ 978-0080999869

• Amazon

Back to Contents