Thursday, December 19, 2013

Carbon CEST and low Spin-lock Field R1rho

Great paper this week by Zhao, Hansen and Zhang from UNC Chapel Hill (& the U of T).  Here is the citation:

The full title is

Characterizing Slow Chemical Exchange in Nucleic Acids by Carbon CEST and Low Spin-lock Field R1rho NMR Spectroscopy.

by Bo Zhao, Alexander Hansen and Qi Zhang

First of all, I've got to give the Zhao et al. credit for a descriptive title.  This JACs communication introduces a new pulse sequence (2D 13C CEST) useful for characterizing slow exchange in nucleic acids.  Then this experiment is compared and contrasted to other tools for characterizing slow exchange, namely R1rho RD and ZZ-exchange.  So as you can see, the title really sums up the paper.

I'll admit that I really like this paper, so let's jump in and see what it is all about.

This paper describes "a nucleic-acid-optimized 2D 13C CEST experiment."  I discussed the concepts behind the CEST experiment a couple of weeks ago, albeit in a slightly different context (see  The experiment described in this week's paper is essentially a selective 2D HSQCs with a long exchange time (Tex) before t1 during which weak B1 is applied to the 13C channel.  This experiment is repeated and the offset of this B1 is arrayed.  The worked-up data (CEST profile) is a plot of the intensity of a given peak vs. offset.  Note that there are as many CEST profiles as peaks in the HSQC spectrum, in principle.  The authors also repeat at different B1 powers.  If the system does not undergo slow chemical exchange the CEST profiles is just a selective saturation experiment (like presat).  When the offset equals the resonance frequency, the peak disappears.  If the system does undergo slow chemical exchange, the saturation of either peak will impact the intensity of the other.  As a result, there are two dips in the CEST profile.

The authors use their new pulse sequence to characterize the ligand-free state of the fluoride riboswitch.  The structure of the ligand-bound form has been determined using crystallography - it is a pseudoknot.  The ligand-free form is hypothesized to have two stem loops based on the 1D 1H NMR.  There are a few extra peaks in the 1D 1H spectrum of the ligand-free RNA, presumably due to slow exchange with another species.  The unstated hypothesis is that this 2nd species is a pseudoknot identical to the ligand-bound form.
One cool trick employed by the authors is to selectively 13C/15N label guanosines in their RNA.  This selective labeling along with a clever use of shaped pulses "to selectively invert and refocus carbon magnetization of interest and to refocus carbon-carbon couplings from neighboring carbons" isolates two G resonances: H8-C8 (base) and H1'-C1' (sugar).  I have played this trick myself, so I know how cool it is to reduce a complicated RNA to a pretty simple and easy to look at spectrum.  If you study the primary sequence of the fluoride riboswitch  you will find 12 G residues.  If you classify these twelve based on the hypothesized secondary structure of the ligand-free RNA and the measured structure of the ligand-bound RNA you would find that - a) there are 4 G residues (G7, G8, G10 and G39) with different secondary structure in free and bound form; b) there are 8 G residues (G1, G2, G4, G14, G23, G30, G31, G33) with the same secondary structure in free and bound form.  These are probes we can use to monitor the slow exchange between these two states using the CEST experiment.

The authors record 13C CEST profiles at 30 C with Tex = 300 ms. (I wonder how they chose these values!).  There is only one dip in the CEST profile of one of the control probes (G33) with the same secondary structure (and presumably chemical shift) in either the ligand-free or bound form.  There are two distinct dips in the CEST profile of two probes (G8 and G10) that transition from a tetraloop secondary structure to a stem-loop upon ligand binding.

The interpretation of this data is that there is an "invisible" excited state (ES) in slow exchange with the ground state (GS) for G8 and G10 (seen in both base and sugar resonances).  For G33 on the other hand, there is no chemical exchange and, thus, no excited state.

The authors make a throw away comment that made me pause: "except for residues from P2, we observed either asymmetrically broadened intensity dips or more than one intensity dip for all other guanosine residues."  I guess this observation makes sense.  After all P1 is a stem-loop in the ligand-free structure and right in the business end of the pseudoloop in the ligand-bound state.  So although it is helical in both structures, it might have different chemical shifts.  So what I am saying is that my division of the 12 guanosines into two easy categories is an oversimplification, at best, and just plain wrong, at worst.

The authors focus their attention on G8 and G10.  Two-state Bloch-McConnell equations are fit to the CEST profiles to extract the ES chemical shift, the exchange rate constant and equilibrium constant.  For the base resonances, the ES chemical shift equals 134.3 and 133.8 ppm for G8 and G10, respectively, which is ~4 ppm upfield from the GS chemical shift.  The average C8 chemical shift for a G residue in a helix equals 133.47.  The exchange rate constant (k1 + k-1) equals 112 +/- 10 Hz.  The population of ES equals 9.8%.  (By the way, with a population that large, why don't you see ES cross-peaks in the HSQC?)  It is trivial to calculate K = 0.112 and deltaG = 5.5 kJ/mol or 1.3 kcal/mol.  Also it is trivial to calculate k1 = 11.2 Hz and k-1 = 100.8 Hz.  (Remember that these values are at 30 C.)

I think it is important to consider these numbers in context.  Arguably, the most prominent NMR dynamics measurement was published WAY BACK in 2005 by Eisenmesser et al. (Nature 2005 438, 117 -  This publication discusses the protein cyclophilin A (CypA).  CypA is in equilibrium between a major and minor form and the authors use relaxation dispersion NMR to measure K = 0.055, k1 = 60 Hz and k-1 = 1080 Hz.  Man, I thought this work was the bee's knees back in the day!  At any rate, this protein refolds with rate constants ~ one-fold larger than the fluorine riboswitch.     

For another example Wenter et al. (Angew. Chem. Int. Ed. 2005 44, 2600 - followed refolding of a bistable 20-mer using real time NMR.  (This paper is one of all-time favorites).  They determine K = 0.236 , k1 = 0.031 +/- 0.006 Hz and k-1 = 0.131 +/- 0.024 Hz at 25 C.  By the way, to ease comparison with the present paper I've flipped the reaction from the Wenter paper so that the more stable fold is the reactant and the less stable fold is the product.  However you look at it, this bistable hairpin refolds with rate constants THREE ORDERS OF MAGNITUDE LESS than the fluoride riboswitch. 

It kind of stuns me that the Wenter 20-mer has such smaller rate constants!  To unfold, the 20-mer does have to break 6 Watson-Crick base pairs and form four new ones.  (I don't know if these two events happen sequentially or simultaneously).  Returning to the Zhao paper, only three or four Watson-Crick base pairs are broken, but nine new base pairs are formed (if we assume the "invisible" species is a pseudoknot identical to the ligand-bound form).  Additionally, some tertiary elements need to form.  So how does the fluoride riboswitch do it so fast?  (This point seems like a good place to admit that I am reading this paper in ASAP form and do not have access to the SI, so I have no idea how much Mg2+ Zhao et al. uses vis-a-vis Wenter et al.)  I guess I should also turn the question around and ask why does the Wenter 20-mer refold so slowly.  Obviously, there is something with the transition state!

To finish up, Zhao and co-workers end their communication by comparing their novel pulse sequence to two more established techniques, low spin-lock field R1rho RD and ZZ exchange.  The results are essentially identical, though ZZ exchange is a bit dicey.  They conclude "The currently presented 2D 13C CEST experiments ... provide powerful tools to investigate slow chemical exchange.  The robustness of these methods promises a unique opportunity to facilitate atomic understanding of slow conformational interconversion that is essential to many vital nucleic acid functions."

I agree.  I like this communication an awful lot and I want to try CEST next time I deal with slow chemical exchange.

Thursday, November 21, 2013


This week we'll be discussing the following recent JACs communication:

A CatalyCEST MRI Contrast Agent That Detects the Enzyme-Catalyzed Creation of a Covalent Bond


Dina Hingorani, Edward Randtke and Mark Pagel

I am a sucker for NMR-based bioanalytical assays.  These experiments appeal to my desire to move NMR beyond its role in structural biology and into other arenas of biology.  Additionally, I have watched these CEST methods with a passing interest, but I have never had a chance to sink my teeth into this kind of work.

In this JACs communication Hingorani et al. describe a MRI contrast agent designed to detect catalysis by the enzyme transglutaminase (TGase).  The authors call this technique CataylCEST.  Here is a rough sketch of how it works.  The authors design a paramagnetic tag with a moiety that looks like the substrate to an enzyme.  (In this case the substrate is lysine and the enzyme is TGase).  Then the enzyme covalently attached the paramagnetic tag to a second substrate (in this case a protein or peptide via glutamine residues).  Once the tag is attached, protons (and 13C, 15N, etc) in the neighborhood to it broaden and resonate at dramatically different frequencies due to hyperfine contact shift.  Let's assume that some of these protons are in slow exchange with the water.  If you saturate these spins using a presat-type experiment, then during the saturation pulse, some of this saturation is transferred to the water by chemical exchange and the integral of the water signal will decrease relative to a control.  Figure 1A of the paper (below) describes this experiment:

The specific enzyme the author's query, TGase, catalyzes covalent bond formation between side chains of glutamine (Q) and lysine (K).  TGase forms cross-links in the extracellular matrix in tissues and in cancer.  The authors do not discuss their motivation for monitoring TGase in the introduction and only touch on it in the conclusion.  I can only surmise that their long-term goal is to develop in vivo CEST MRI with exogenous CEST agents that do not suffer from rapid pharmacokinetic washout.  

The results of this study - 

The author's synthesize TM-DO3A-cadaverine paramagnetic CEST agent and couple this molecule to 5 different substrates using inexpensive microbial TGase.  The substrates are: Boc-Gln-OH, CBZ-protected QG peptide, QR peptide, GQR peptide and bovine serum albumin (BSA).  They acquire their NMR data on a 600 MHz NMR at 37 C using (essentially) the presat experiment with a saturation time of 4 s and a field of 20 uT (or (20 uT/14.1 T)*600 MHz = 850 Hz field).  The saturation offset is arrayed in 1 ppm steps from -30 to 30 ppm, relative to water, which is set at 0 ppm.  The data is presented as a saturation profile showing the intensity of the water signal as a function of the saturation offset.  Finally, the authors fit Lorentzian line shapes to assess the chemical shift of the species in slow exchange with water.  The authors do not report error bars, which I find troubling, because the peaks are quite broad.  As we'll see later, the authors will argue that a difference of 1.8 ppm is significant, but a difference of 1 ppm is not.  I am not sure what to make of this difference, myself.

The most convincing result is with BSA.  The authors make an NMR sample with 25 mM CEST agent, 0.75 mM BSA and 10 mM glutathione (to maintain the reducing environment required by the eznyme) in pH 7 Tris buffer.  Then they react with 0.327 uM TGase for 24 h under aerobic conditions.  The authors repeat this experiment in triplicate.  Before catalysis, the saturation profile shows CEST at +4.6 ppm upfield from water.  After cataysis, the saturation profiles shows a CEST at +4.6 ppm and -9.2 ppm.

The interpretation is that the signal at +4.6 ppm is diamagnetic amines and amides in BSA.  TGase catalyzes the formation of a covalent bond between the CEST agent and Q side chains on BSA.  Hyperfine contact shift of Tm(III) induces an upfield shift in the chemical shift of the amides to -9.2 ppm.  (By the way - do you notice how much easier it is to look at the Lorentzian fits on the right than the saturation profiles on the left!)

If the publication ended right here, I'd be sold.  This paper confuses me, though, when the authors report their controls.  Their results reinforce a long held suspicion about CEST: how do you assign the peaks in your saturation profiles?

Let's discuss their controls.  The most logical control is each of the components individually.  The CEST agent alone shows a CEST at -5.2 ppm, BSA alone shows a CEST at +5.6 ppm and glutathione alone shows a CEST at +5.4 ppm.  By the way, I'll note that the text says that BSA and glutathione have a CEST at +4.6 ppm, but the figure caption says +5.6 and +5.4 ppm, respectively.  You see what I mean about error bars!  I guess 1 ppm is not significant.  So why isn't the CEST profile of the reaction mixture before catalysis equal to a superposition of the reactants?  (I think it probably is, if you correct for concentration, but the authors do not do this for the readers and we are left wondering).  Next the authors try control peptides.  The authors mix 25 mM CEST agent, 25 mM peptide, 10 mM glutathione in pH 7 Tris buffer with and without TGase.  For GQR and QR, there is a CEST at -9.0 ppm before catalysis and CESTs at +5.8 ppm and -10.8 ppm after.  (In the discussion, the authors assign the signals at -9 ppm and -10.8 ppm to a supramolecular adduct and paramagnetic amides, respectively, implying that they can tell a difference between signals the differ by 1.8 ppm.  See what I mean about error bars!)  For CBZ-protected QG, there is no CEST signal before catalysis and CESTs at 4.6 ppm, 9.8 ppm and 22.5 ppm after.  Finally for Boc-protected Q there is a CEST at +7.2 ppm before and a broad CEST between -10 ppm and -20 ppm after catalysis.  Why do none of the controls show the same signal before catalysis as BSA?  (Once again, I presume the issue is concentration, but the authors leave it up to their reader to discern).  Also, after covalent attachment of the CEST agent why does the CEST differ depending so dramatically for different substrates?    

The author's interpretation of their results falls into two categories (they don't make this explicit, I am interpreting).  1) Aggregation/Heterogeneity - There is a noncovalent supramolecular adduct or some type of conformational heterogeneity that dramatically alters the chemical shift of the water exchangeable protons.   This effect is responsible for CESTs at -5.2 for the CEST agent alone, at -9 ppm for GQR and QR peptides prior to TGase catalysis, at +4.6 ppm, +9.8 ppm and +22.5 ppm for the CEST agent-linked ZQG peptide and at -10 to -20 ppm for the CEST agent-linked Boc-Gln-OH.  2) Change in chemical exchange rate contants -  The authors assert that chemically modifying the substrate alters the rate constant and rates of chemical exchange.  This effect is responsible for the CEST at +5.8 ppm for GQR and QR peptides after TGase reaction.

As you can tell, I am skeptical of these explanation.  I am not saying the authors are incorrect.  They know a lot more about this subject than I do.  I only mean to say that I believe more justification is needed.  For example, there are a few additional controls the authors could do to validate their assignments.  If they are concerned about "noncovalent supramolecular adducts", why not reduce concentration or increase ionic strength to break up these interactions?  If they are concerned about hydrophobicity causing conformational heterogeneity or heterogeneous ligand conformations, then perhaps the peptides they are using are not good controls?  If they are concerned about rate constants, why not play with temperature or field?  

In the end, this paper DOES convince me of its main objective: the authors have a catalyCEST MRI contrast agent that can be used to detect the formation of a covalent bond in BSA mediated by TGase.  After studying this paper, though, I am concerned that the CEST varies from substrate to substrate - from -9.2 ppm for BSA to 22.5 for ZQG.  Let me end my critique with a question: If you wanted to check for TGase activity using a new protein and you were not sure if it was a substrate or if you wanted to test TGase activity in vivo, then what results do you expect from the CEST assay?  The fact that you do not know shows how far we have to go. 

Thursday, November 14, 2013

Dissecting the stereocontrol ...

My blog post this week will critique the paper:

Dissecting the Stereocontrol Elements of a Catalytic Asymmetric Chloroactonization: Syn Addition Obviates Bridging Chloronium.


Yousefi, Ashtekar, Whitehead, Jackson and Borhan

Funny story about this JACs communication (JACs 2013 135, 14524): I was reading this paper while riding a stationary bike at my campus gym, when a certain faculty member approached (name withheld to protect the guilty).  He asked "What are you reading?".  I showed him the paper and he said "That looks brutal!"  I laughed.  This science in this paper is excellent, but it was a poor choice by me for a blog that focuses on NMR.  By the time I realized that the NMR methodology was not novel, I was too deep into the paper to pull back.  Nevertheless, I think this publication is worth discussing, because it presents an interesting solution to a difficult diastereospecific assignment problem.

As an exercise for myself, I'll see if I can briefly describe the problem and why this paper is relevant to people interested in NMR.  Then I'll show the author's data and interpretations.

Yousefi et al. are interested in the reaction below: an enantioselective halocyclizations of alkenes.

This type of chemistry is important as "a robust, versatile route a wide range of heterocycles."  The specific question that the authors address is the following:  Why is this reaction enantioselective?  To address this question, the author use NMR spectroscopy of isotopically labeled precursors to assess the mechanism.  Figure 1 describes the two proposed intermediates: path a - three-membered chlorenium ion intermediate; path b - carbocation intermediate.  SPOILER ALERT (this part is in the title): It is path b.

These two mechanism offer some predicable consequence.  If the reaction proceeds by path a, the "enantioselectivity would be controlled in parallel with the initial asymmetric chlorenium delivery, yielding anti addition".  If the reaction proceeds by path b, "the reaction's enantioselectivity would be determined at the (presumably catalyst controlled) ring-closing step."  If the author's can show syn addition, then path a can be ruled out.  The problem demonstrating syn addition is that "the chlorine resides on a nonstereogenic carbon with no record of its attach path on the alkene."  So it is not easy to determine the addition, unless you can somehow label the product!    

The authors synthesize a E-deuterated analog of the alkene.

Then they perform the halocyclization reaction and analyze the products by NMR.

At this point I'll note that the publication itself has ZERO NMR spectra in any figure.  So you'll have to dig through the SI to see any of their data.  Here is the spectrum of their products.

No integrals or peaks picked, but I see 2 methylenes from the lactone at 2.5 and 2.8 ppm and the 5 phenyl protons at 7.4 ppm.  At 3.75 ppm is the CH adjacent to the Cl.  To remind you, below is the expected molecule for the protonated starting material:

One relevant factor to consider is that their deuterated analog is not perfect.  They report 94:6  E:Z and 88% D-incorporation.  So they have a bit of protonated product in with their deuterated product.  Looking at this data I see a "roofed" non-first order CH2 from the protonated product (peaks at 3.815, 3.805 and 3.772 ppm with the final leg hidden under the big peak at 3.74 ppm) overlapping with two CHD diastereomers at 3.805 and 3.74 ppm. 

The author's interpretation is explained in the following figure from the SI:

You may be asking how they know that c is the R,R or S,S diastereomer and f is the R,S or S,R diastereomer?  This gets to the heart of the difficult problem in diastereospecific resonance assignment.  Let me quote directly from the paper: "The absolute stereochemistry of the CHDCl group in the major isolate was straightforwardly established via NOE analysis of epoxide 3-D obtained from the chemically transformed chloroactone product 2-D"

To translate, the deuteratd epoxide (3-D) has a peak at 2.98 ppm in 1D 1H NMR spectrum.  The protonated epoxide (3) has two peaks at 2.98 and 2.73 ppm.  Using NOE, the peak at 2.98 is established as trans the phenyl group.  (This data is nowhere to be found in the body of the paper or SI, by the way.)  Hence the deuteron in 3-D must be cis.  Retrosynthetic analysis can be used to establish the stereochemistry of the major product 2-D.

I'll leave it to the organic chemists to fight through their mechanistic arguments.  Their NMR argument is interesting to me, though.  Basically, the authors cannot complete diastereospecific assignment of 2-D using conventional approaches, so they modify the molecule in a stereochemically predictable manner to simplify the problem.  I find this approach to be clever.

Like I said at the beginning, this article is not for the faint of heart.  I don't know that I am really all that interested in mechanistic details of this reaction.  My interest is in the NMR.  Frankly, the authors do not wow me with their NMR, but I am intrigued by their solution to the diastereospecific assignment problem.  

Thursday, October 31, 2013

NMR Chemosensing

Welcome to my new NMR blog.  I hope to use this blog to discuss interesting new publications in the literature.

Today I'm going to review a recent JACs communication by Perrone et al.

The title of this paper is '"NMR Chemosensing" Using Monolayer-Protected Nanoparticles as Receptors.' 

A chemosensor is a "molecule of abiotic origin that signals the presence of matter or energy".  As a very general concept, a chemosensor can be imagined as a "supramolecular receptor" that selectively binds its target (the analyte), which causes some measurable change in a physical property (such as absorbance, redox potential, relaxivity, etc.)  It does not take a vivid imagination to dream up uses for chemosensors and thus, there are many scientists in the world making chemosensors.  To distinguish their chemosensor from all the other sensors out their, Perrone and co-workers make the following observation regarding the current state of chemosensor research:  for most chemosensors "the signal produced arise from a property of the chemosensor itself and therefore does not provide any direct information on the identity of the analyte detected."  These authors approach is the opposite.  In this publication they describe a chemosensor in which the analyte-sensor interaction alters a physical property of the analyte.  The measurement step queries physical properties of the analyte not the sensor.  OK.  It is novel.  Is it useful?  (I am not 100% convinced that the author's approach offers advantages over the more common receptor-centric approach, but I'm willing to give them the benefit of the doubt at this stage.)

How do they makes these sensors?  They use 2nm diameter gold nanoparticles coated with thiol monolayer, which seems to be the expertise of this group.  This monolayer is amphiphilic (for example, 7 carbon alkyl chain linked to a triethylene glycol head group via amide, 1).

Next, a mixture containing 70 uM chemosensor and 7 mM of 3-5 water soluble organic acids or sulfates (see the roster below) in D2O is prepared.   

A curious choice of molecules!  I wonder how the authors settled on these five.   

At any rate, let's get down to the experiment  It is expected that hydrophobic molecules will partition into the hydrophobic moiety of the monolayer, whereas hydrophilic molecules will interact with both the polar headgroup and polar solvent.  This mixture is probed using a NOE-pumping experiments, that works like STD NMR.  Briefly, the nuclear spins of aliphatic region of the monolayer are perturbed (inverted or saturated).  Some of this perturbation is transferred via NOE to small molecules partitioned into the hydrophobic monolayer.  Assuming these molecules are in fast exchange with free solvent, there will be signals in the resulting spectrum that arise from the analyte (not the sensor) as a direct consequence of interaction with the sensor.  The fact that the analyte has to be in fast exchange with the receptor has consequences for the detection limit, acquisition time and affinity towards the receptor.   

The first example presented is a mixture of 4, 7 and 8.  The chemosensor + NOE pumping experiment shows remarkable selectivity and signals from only one molecule are detected! 

The top spectrum is the aromatic region of the 1D 1H of a mixture of 4, 7 and 8 at a concentration of 7 mM in D2O.  The bottom spectrum is the aromatic region of the NOE-pumping experiment of the same mixture with 70 uM 1-coated gold nanoparticles.  I am impressed.  Two comments, though.  First, you can almost see a little bit of 7 peaking out of the noise.  Am I wrong?  Second, the authors do not mention the experimental conditions of the NOE-pumping experiment.  I understand that this publication is a JACs communication and the authors do not have a lot of space for extraneous information.  It is just that I'm curious about the sensitivity of the NOE-pumping experiment.  How many scans did they have to record to get the bottom spectrum?  (At a later point in the paper it is implied that the acquisition time is 4 hours using NOE-pumping.)

The authors follow up on this experiment by demonstrating the selectivity of their chemosensor.  For instance, in a mixture of three isomers of salicylate (4, 5 and 6) with 1-coated nanoparticles, the NOE-pumping experiment only show signals for 4.  (This data can be found in the SI.  I have no idea why the author's buried it there.)  Once again I am impressed, even if a bit of 5 and 6 seem to peak out of the noise.  The authors measure the apparent association constant, as well, to demonstrate that quantitative analyte detection with their chemosensor.  The detection limit for salicylate is 2.5 mM (OK, now I am a bit less impressed.  By the way, at the end of the paper the author's mention STD-NMR can be used in lieu of NOE-pumping.  Using STD-NMR the detection limit is 250 uM and the acquistion time is 30 minutes.  Why did they bother with NOE-pumping?)  At this stage the authors have convinced the reader that their 1-coated nanoparticle in combination with the NOE-pumping experiment is selective for sodium salicylate, even in the presence of isomers. The authors do not report results for the detection of salicylate (4) in a solution enriched in 5 and/or 6.  One might accuse them of dodging the issue of false positives, but I'll cut them some slack because this article is a JACs communication.  The authors are excellent communicators and do a nice job of leading their audience away from the short-comings by highlighting the positives.  I wish I were this skillful at this technique!

Since the authors have convinced us of the "what", they turn the "how" and "why".  Why do 1-coated nanoparticles and the NOE-pumping experiment detect 4 and not 5?  It goes back my introduction of their technique.  Hydrophobic molecules partition themselves into the aliphatic moiety of 1.  The authors note that the elution order of a mixture of 4-8 using reverse-phase C18 HPLC is 8 << 7 < 6 < 5 << 4.  The interpretation of this result is that 4 is the most hydrophobic.  Likewise the computationally predicted logD at pH 7.4 concurs with the elution order.  In other words, it seems salicylate really is more hydrophobic than the isomers and the chemosensing technique described in this publication relies of this fact.  Of course, this highlights a potential weakness in selectivity.  Will you ever be able to distinguish between two hydrophobic compounds (with equal C18 retention time and/or logD)?  Can we modify the chemistry of the coating thiols to distinguish molecules using a property other than hydrophobicity?

The authors put a positive spin on these questions.  Let me quote the paper.  "Since there are no limitations on the chemical structures of the analytes and the nano particle-coating thiols, the NOE pumping experiment has the advantage of very general applicability."  Really?  No limitations?  The authors make 2 and 3-coated nanoparticles and return to their mixture of 4, 7 and 8 at a concentration of 7 mM in D2O.  The aromatic region of the NOE-pumping experiment with 2-coated nanoparticles as the receptor shows signals from 4, 7, 8 and an impurity of 8 (middle spectrum, below).  The authors refer to this result as "different selectivity."  I would call this NO selectivity myself, but hey you say PO-TAY-TOE, I say PO-TOT-TOE.  (  The aromatic region of the NOE-pumping experiment with 3-coated nanoparticles as the receptor shows signals from 4 and 7 (bottom spectrum, below).

Finally, to demonstrate applicability the authors spike 5 mM salicylate into a sample of human urine, mimicking the concentrations found in urine after a medium-dose administration of acetylsalicylic acid (aspirin).  The aromatic region of the 1D 1H spectrum is complicated due to the myriad of metabolites in urine.  The aromatic region of the NOE-pumping spectrum of the same sample in the presence of 1-coated nanoparticles show only signals for salicylate.   

Let me conclude by stating that I do not want my criticisms to overshadow the excellent science in this communication by Perrone et al.  I am a nobody, whereas the authors are doing the hard, creative work of producing scientific knowledge.  I am genuinely impressed by their approach and I appreciate that this publication is the first step in a long journey.  Having said that I will be more impressed when they design a sensor to detect something a bit more interesting than 5 mM salicylate spiked into urine!  One example that springs to mind is enantiomeric impurities.  There are pharmaceuticals (for examaple, penicillamine - that have a single chiral carbon.  One enantiomer is a powerful therapeutic agent; the other is a deadly poison!  In 2004 Tsourkas and co-workers published a communication ( describing an antibody-based magnetic relaxation switch capable of detecting an impurity of 0.1 uM D-Phenylalanine in the presence of 10 mM L-Phenylalanine.  (That is 99.998% ee if you are keeping score at home!)  It is difficult for me to see how the monolayer-protected nanoparticles introduced by Perrone et al. will approach this level of sensitivity and discrimination.