Wednesday, October 29, 2014

Electron Paramagnetic Resonance of hair!

I wanted to switch gears a bit and do a paper on Electron Paramagnetic Resonance (EPR), also know as Electron Spin Resonance (ESR).  The paper for this weeks is

Electron spin resonance (ESR/EPR) of free radicals observed in human red hair: a new, simple empirical method of determination of pheomelanin/eumelanin ratio in hair.


Chikvaidze EN, Partskhaladze TM and Gogoladze TV;jsessionid=39B010DF34D055C5F27880DD49E628F1.f01t04

I discussed EPR briefly in an earlier post (  To review, one can think about EPR much like you think about NMR.  In a simple 1D 1H NMR spectrum parameters like the resonance frequency (the chemical shift plus any scalar couplings) and peak integrals can be interpreted to understand properties of the molecule, for instance the molecular structure of a small drug-like organic molecule.  There are similar parameters for EPR.  In EPR the resonance frequency is reported as the "g-factor."  Instead of depending on the shielding of nuclei by electrons like the chemical shift in NMR, the g-factor depends on the coupling of the spin motion of the electron to the orbital motion (spin-orbit coupling).  These days every organic chemistry textbook contains a table of chemical shifts classified by functional group.  With EPR, on the other hand, I am not aware of any standard tables of g-factors.  I don't want to suggest that the spin-orbit coupling is not sensitive to electronic structure.  Of course it is!  Spin-orbit coupling depends on which orbital the unpaired electron resides.  For metals, the g-factor is crucial.  For organic (oxygen, carbon and nitrogen) radicals, though, it seems like all g-factors are ~ 2.  Although the g-factor is not diagnostic in these cases, the coupling to magnetic nuclei are!  The unpaired electron will experience "hyperfine" coupling to nuclei, such as 14N (nuclear spin I = 1).  The electron is split into 2I+1 lines.  Hence if the nuclei is 15N (I = 1/2) the EPR signal is a doublet.  If the nuclei is 14N, the EPR signal is a triplet, but all three legs have identical height.  (As an aside, I'll mention to organic chemists that they should look at the CDCl3 signal in a 1D 13C spectrum to see an equivalent effect).  EPR spectra differ in two key ways from NMR spectra.  First, the x-axis is magnetic field (in gauss or telsa) not frequency.  Second, the y-axis is the first derivative of the absorbance.

Now that we are all experts on EPR, what is it that Chikvaidze and co-workers are measuring?  There is a branched polymer in skin and hair called melanin that controls pigmentation color.  This polymer is made of varying amounts of two monomers, called eumelanin and pheomelanin.  The former is associated with dark (brown or black) colors, the latter with red.  Eumelanin contains a O-C-C-O semiquinone and gives an EPR signal consistent with an oxygen radical (a singlet).  Pheomelanin contains a O-C-C-N semiquinonimine and gives and EPR signal consistent with a nitrogen radical (triplet).  According to the authors the measurement of the concentration of pheomelanin in skin samples is "an issue of great interest in the world" because UV-mediated breakdown of this molecule produces reactive oxygen species which "might help to explain the relatively high incidence of skin cancer among red-haired individuals."  There are assays available to determine the ratio of pheomelanin/eumelanin in hair samples that involve chemical treatments, etc.  Because these molecules are paramagnetic, it is possible to use EPR as a non-invasive assay to determine the amount of pheomelanin (ug/mg) in hair samples.

What did the authors do?  They collected 113 hair samples from their students (42 black, 28 dark brown, 27 red and 16 blond - I assume each sample is from a different student.  It is not clear if all samples are used in this study.) and divided the samples in bundles of equal mass (40 mg) and length (1.5 cm).  Because they need to make very accurate measurements of the g-factor, the authors use a standard of Mn2+ (in MgO powder).  I assume the standard is in a sealed capillary which is placed in a 4 mm EPR tube along with the hair bundle.  EPR are measured using an X-band EPR spectrometer at room temperature.

The EPR spectrum of black hair is a "slightly asymmetric singlet" with g-factor = 2.0035-2.0036.  The spectrum of black hair looks identical for each donor (data not shown).  For red hair, on the other hand, the spectrum vary depending on the donor.  The authors classify two distinct types of spectra (type I and type II) with g-factor = 2.0038-2.0047 shown in Figure 2:

The interpretation of this data is that the EPR spectrum of black hair is essentially the spectrum of eumelanin, whereas the spectrum of red hair is a superposition of eumelanin and pheomelanin.

The authors used microwave saturation to filter one of the components of red hair.  Let me try to explain saturation succinctly.  If the rate of spin flips between the spin-down and spin-up state is faster than the rate of relaxation back to the ground (spin-down) state, then the intensity of the signal is attenuated by saturation.  The rate of spin flips depends on microwave power.  The relaxation rate depends on R1 (= 1/T1).  Let's say there are two signals in the EPR spectrum.  One has a small R1 (aka large T1, aka slow relaxation) and the other has a large R1 (aka small T1, aka fast relaxation).  We can then choose a microwave power such that the rate of spin flips is larger than the small R1, but smaller than the large R1.  So the slow relaxing signal is saturated and maybe even disappears.  One can play this game with a 10 uM solution of TEMPOL radical under an atmosphere of air (which includes paramagnetic oxygen that will increase R1) or nitrogen.  At high microwave power the slow relaxing (large T1) nitrogen atmosphere sample will saturate and give no signal, but the fast relaxing (small T1) air atmosphere sample will not.

Figure 3 shows the EPR spectrum of red hair as a function of microwave power:

The authors interpretation is that "the triplet (hyperfine coupling = 0.372 mT, g-factor = 2.0055) .. evident after saturation of the singlet at maximum microwave power, corresponds to pure pheomelanin in the .. spectrum."

Now the authors are prepared to address their main goal: determine the ratio of pheomelanin/eumelanin in hair.  Given that pheomelanin has a g-factor of 2.0055 and eumelanin has a g-factor of ~2.0036, the authors hope to use the experimental g-factor of hair to determine the ratio.  Normally, one would make a calibration curve with different ratios to test the fitness of this method. Not so simple in this case.  One can mix different ratios of red and black hair, if you assume black hair is pure eumelanin.  Figure 4 shows the calibration curve.

Using data from Ito et al. (Pigm. Cell Melanoma Res. 2011 24, 605) that correlates hair color to concentration of pheomelanin in ug/mg, the authors make another calibration curve to convert the measured g-factor into pheomelanin concentration, shown in Figure 7.


My only beef with this paper is that the accuracy and precision of this method to measure pheomelanin and the ratio of pheomelanin:eumelanin is not addressed.  Let me state it another way.  Let's assume you can distinguish hair that contains 0.8 ug/mg pheomelanin from hair that contains 4.8 ug/mg.  Can you distinguish 2.4 ug/mg from 2.5 ug/mg?  To answer this question I did some simulations in MatLab using EasySpin ( a supercool simulation program.  My script is at the end of the blog.

The first plot shows the simulated EPR spectra of blond (in blue) vs red (in green) hair using data from Table 2 in the paper. 

Using the zero crossing to estimate the g-factor for the blond hair equals 2.0035 and for the red hair equals 2.0037, which is somewhat smaller than the 2.0046 from the Table 2 of the paper.  OK fair enough.  We can tell blond from red hair using the g-factor.

Two different red hair samples with pheo/eu% equal to 59 (blue) and 54 (green) look identical to me (see below).


Maybe I'm screwing something up with my simulations, but based on these results I am skeptical that X-band EPR can pick up subtle difference is pheomelanin:eumelanin.  Q-band (35 GHz) is a different story, though.  Signals at different g-factors no longer overlap and using double integrals, it may be possible to estimate pheomelanin concentration.  Below is the Q-band simulation of pheo/eu% equal to 59 (blue) and 54 (green).  I also assumed smaller lwpp to exaggerate the transitions.

Overall, I like this paper because it made me think about EPR and play with EasySpin.  It also got me worried about hair in my EPR cavity!  I'll have to keep the gingers away from my instrument.


EasySpin MatLab script

% Mixture of pheomelanin and eumelanin
% based on data from Chikvaidze et al. MRC 2014


% Experimental parameters
Xp.mwFreq = 9.43;
Xp.Range = [334 338];

% Component 1
Eu.g = 2.0035;
Eu.lwpp = 0.5;

% Component 2
Pheo.g = 2.0055;
A = 0.372;
Pheo.A = mt2mhz(A);
Pheo.Nucs = 'N';
Pheo.lwpp = 0.5;

% Relative abundances
Eu.weight = 0.46;
Pheo.weight = 0.54;

% One call to pepper
[B1,spc1] = pepper({Eu,Pheo},Xp);

% Relative abundances
Eu.weight = 0.41;
Pheo.weight = 0.59;

% One call to pepper
[B2,spc2] = pepper({Eu,Pheo},Xp);

plot(B1,spc1,B2,spc2); Chikvaidze, E., Partskhaladze, T., & Gogoladze, T. (2014). Electron spin resonance (ESR/EPR) of free radicals observed in human red hair: a new, simple empirical method of determination of pheomelanin/eumelanin ratio in hair Magnetic Resonance in Chemistry, 52 (7), 377-382 DOI: 10.1002/mrc.4075

Monday, October 13, 2014

You don't need that big expensive magnet to do NMR!!!

As chemists we often focus on parameters like chemical shift, scalar coupling and integrals that can be measured in an NMR spectrum and interpreted to understand qualitative and quantitative information about molecules.  There are applications where these parameters are less important and the longitudinal (T1) and transverse (T2) relaxation time constants and/or the self-diffusion coefficient (D) are critical.  A lot of these applications do not involve the types of samples that organic chemists or biochemists prepare (~600 uL in a 5 mm NMR tube).  In fact, sometimes T1, T2 and D need to be measured in extreme environmental conditions.  You can't drag your 11.7 T magnet to the south pole!  Even in more benign environments, a big magnet is not necessary for T1, T2 and D measurements needed for tasks like food characterization and oil-well logging.

Today I am going to discuss two papers that explore NMR without a big and expensive magnet!

The first paper is

"Ultra-low-field NMR relaxation and diffusion measurements using an optical magnetometer"


Ganssle PJ, Shin HD, Seltzer SJ, Bajaj VS, Ledbetter MP, Budker D, Knappe S, Kitching J and Pines A.

Angew Chem Int Ed Engl. 2014 53 9766-70
doi: 10.1002/anie.201403416

The authors design and demonstrate an ultra low field (ULF) NMR capable of performing industrially relevant measurement (T1, T2 and D) for the characterization of mixtures of hydrocarbons and water.  The authors claim that their instrument is the first step towards a compact, inexpensive and robust NMR sensor which operates at the Earth's magnetic field. 

How does this work?

Conventional NMR detectors use a coil to detect transverse magnetization.  The ULF NMR uses a magnetometer like what is in your cell phone as a compass.  The specific magnetometer is optically detected (the author's do not really explain the detector in this paper).  I don't pretend to understand all the details of this detector, which is called a spin exchange relaxation-free (SERF) configuration magnetometer.  The important thing to understand is that the detector measures longitudinal magnetization (along the z direction), in contrast with traditional NMR coils.  In fact, during acquisition, a series of 180 degree pulses are applied to sample to flip the spins between the +z and -z direction.  The authors convert the "average magnitude of change in magnetometer signal in response to a pi pulse" into a sensible signal.   

The authors want to design "a NMR sensor which operates at the Earth's magnetic field", but for now they have to make a few compromises to get a prototype.  First, there isn't much longitudinal magnetization in samples polarized by the Earth's magnetic field, so the authors apply a "pre-polarizing" 2 T field.  Second, the sample chamber is not really at the Earth's magnetic field, which (according to Wikipedia is at 25–65 uT (microtesla).  By the way, a strong refrigerator magnet has a field of about 10 mT.  At any rate, the chamber is designed to have no magnetic fields - there is some sort of special shielding to remove the Earth's magnetic field.  Then a weak "bias" magnetic field (50 uT in this case) is applied to the chamber.  This field is designed to be turned off during detection (the SERF detector has optimal response at zero field) and during pulses.  The reason for turning off the field during pulses is so that "the device does not need to be re-tuned when the bias field is changed."   Overall, their set up looks like the following:

The pulse sequences used to measure T1, T2 and D are the following:

If you study these pulse sequences, they will make a lot of sense.  The T1 measurement does not use the inversion recovery pulse sequence, instead longitudinal relaxation time constant is measured "as an exponential decay of the spin magnetization as a function of an increasing delay time between the sample pre-polarization and measurement."  You also see the unusual detections strategy.  Longitudinal magnetization is flipped between -z and z by a pi pulse.  The T2 measurement is a standard CPMG train.  "The inter-pulse spacing (tau) is held constant and the number of pulses (n) is varied."  Diffusion is CPMG with a gradient.

One thing I find pretty cool about this set up is that it uses conventional 5 mm NMR tubes!

What do the authors measure?

They record the relaxation time constants (T1 and T2) of some common solvents: water, methanol, ethanol and hydrocarbons at 0.5 G and 37 C.  They also make mixtures of hydrocarbons and water.  How did they do that?  These liquids are immiscible!  In fact, they used "a coaxial insert which separated liquid in a smaller, 3.3 mm NMR tube from the liquid in a standard thin walled 5 mm NMR tube."  I'm not sure I would call that a mixture, but why split hairs!  The data looks like the following:

Finally the authors can make 2D plots of T2 v T1 or T2 v D to demonstrate how readily these solvents can be distinguished based on these relaxation properties:

The authors conclude that this device is an "important first step towards the development of compact, inexpensive devices which can take advantage of the Earth's high homogeneous ambient magnetic field."  They acknowledge "the limitations present in these experiment ... are artifacts of the design of the device" but feel that they make a "compelling case for future research".

I think the device described in this publication is interesting and important.  I have a few questions, though.  How long does it take to make each T1, T2 and D measurement shown in figure 3 (each dot on the curve)?  Can T1, T2 and D distinguish miscible liquids, such as water-methanol or water-ethanol?  I work at a University and the proof of the vodka is a major concern on football days!  Finally, how big is the instrument, really?  It is hard for me to get a feel for the size based on Figure 1.  It looks almost as big as a superconducting NMR magnet.

The second paper is

"Scalable NMR spectroscopy with semiconductor chips"


Dongwan Ha, Jeffrey Paulsen, Nan Sun, Yi-Qiao Song and Donhee Ham

In contrast with the first paper, which described an unconventional detection scheme, this paper describes a miniaturized NMR system that is ~12 cm^3, weighs 7.3 kg and can perform 1D 1H, multi-pulse and heteronuclear experiments.  By contrast, a commercially available system like the Picospin45 ( is 20.3 cm x 14.6 x 29.2 cm, weighs 4.7 kg and can only do 1D 1H.  The major innovation by the authors is the miniaturization of the electronics into a 4 mm^2 chip.  The technical details are best understood by an electrical engineer, so I won't explain too much.  The net result is a system that looks like the following:

Obviously there is quite a bit missing from the picture, but the penny on the left gives a general idea of the size of the system.  The Larmor frequency is 21.8 MHz.  Samples are 1 mm capillaries with ~0.8 uL of sample.  What can the system do?  Figure 3 shows 1D 1H for 7 small molecules (acquisition times on the order of seconds to minutes).

Frankly, this data looks like 21.8 MHz NMR spectra.  The magnet is shimmed to 0.13 ppm resolution (~2 Hz).  So you can make out 7 Hz couplings, but small couplings or lines closer than ~0.1 ppm blend together.  The spectra of aspirin, serine and glucose are not useful for chemical characterization.

One of the advantages of the design is pulse programming.  The ability to control pulses and delays enables the authors to go beyond traditional 1Ds and do multidimensional NMR.  Figure 4 shows the JRES and 2D phase sensitive COSY on neat ethanol and 1.5 M alanine dissolved in D2O (acquisition times are 15 and 73 minutes, respectively).

The authors can also collect HSQC and HMQC on 13C enriched methanol in 17 and 34 minutes, respectively.  Note that there is no decoupling during acquisition (t2), meaning the peaks are split by the 1JCH.

The authors round the paper out with a relaxometry experiment on a crude oil sample.  They also introduce a clever processing hack to handle temperature fluctuations, which present a "significant obstacle towards portable NMR."  Their solution deserves a longer explanation in this blog post, but lets face it, it is getting way too long.


To wrap up this post, I'll note that I am not convinced that either system discussed above or a commercial benchtop NMR (from PicoSpin, Magritek or Nanalysis) can ever replace a trusty 400-600 MHz NMR for resonance assignment in organic chemistry or biochemistry.  I will concede that not all chemical characterization requires a 400-600 MHz NMR.  To paraphrase John Edwards from Process NMR associates at the MestreNova Users meeting prior to the ENC this year "if your spectrum looks like crap at 400 MHz, it won't look too much worse at 90 MHz, so why waste time on a superconducting magnet."  The authors of the papers I reviewed today present two novel NMResque instruments capable of making measurements nearly identical to a high field superconducting system.  The trick will be to continue to develop these systems and find applications where these systems outperform conventional systems.  


Ganssle, P., Shin, H., Seltzer, S., Bajaj, V., Ledbetter, M., Budker, D., Knappe, S., Kitching, J., & Pines, A. (2014). Ultra-Low-Field NMR Relaxation and Diffusion Measurements Using an Optical Magnetometer Angewandte Chemie International Edition, 53 (37), 9766-9770 DOI: 10.1002/anie.201403416  

Ha, D., Paulsen, J., Sun, N., Song, Y., & Ham, D. (2014). Scalable NMR spectroscopy with semiconductor chips Proceedings of the National Academy of Sciences, 111 (33), 11955-11960 DOI: 10.1073/pnas.1402015111