Thank You misty yes we're going to talk about NMR we do have NMR capabilities at the University of Michigan

and there are several investigators who are using NMR metabolomic so I'm going to talk a little bit about NMR
very briefly just a very brief overview of its application and a little bit about the capabilities of NMR to detect and
quantify metabolites but I want to primarily focus on some of the examples of the some of the work that we're
doing in particular how NMR metabolomic can complement and be additive to some of the other analytical
platforms in particular lc-ms so NMR is one of several analytical platforms that can be used for metabolomic
you've heard a lot about GCMs and LCMS over the last few days and like LCMS virtually any sample type can
be assayed by NMR from metabolomic LCMS is more sensitive than NMR we detect in the micromolar range
and depending on the sample type we detect and quantify about 40 to 70 samples or sorry metabolites in a
sample unlike LCMS NMR is not destructive to the stand of the sample its quantitative it's highly reproducible
and you generate all your data from a single run so there are some potential advantages of NMR so I am full
full declaration I am NOT an NMR spectra spectroscopy but I do know one who's here today if you have any
specific questions about NMR letters say Yeomans is here who runs the NMR in the College of Pharmacy but
the basic principle behind NMR is uh is it is a magnetic field generating a radiofrequency input to the sample
here and typically we use proton NMR so it causes the protons to resonate and that is detected through
another radio frequency output and that ultimately leads to the generation of a speck from and historically NMR
has been used very often and has a long history of single compound measurements and structural
measurements but essentially we're exploiting that capability to measure multiple metabolites in a biological
mixture like blood and it is able to do that because virtually every proton containing compound has a unique
NMR signature I want to say just a little bit about blood collection I'm going to focus on that as an example of a
bio fluid that we often use for NMR metabolomics because there are some subtle differences between the
sample collection procedures for NMR versus lc-ms there are similarities one is that for serum because there's
no preservative in the use of serum it's actually an ideal sample to use for cross-platform because LCMS thank
you guys like EDTA and the NMR side we like heparin so for serum it's an example of a type of sample that
could be used very easily across both both platforms for NMR in particular we like to preserve blood with
heparin and we like sodium heparin and just keep in mind that the green top tubes that are used in the clinic
are not necessarily endotoxin free and so we advise that you immediately place them in an ice water bath and
show them quickly and freeze them at minus 80 upon collection either generating plasma or whole blood
however you want to do that another difference between NMR is that we do require a larger sample volume
ideally about five hundred microliters we can work with smaller volumes and I'll touch on that to show you how
we can do that but that is our ideal volume and of course we advise as I think Anna mentioned in her talk to
generate not only biological replicates but technical replicates Oh as well so just continuing on with blood as an
example for NMR you can get away with getting metabolite data from an unexpected sample and in this case
um we this is a way you can get a get with smaller volumes if that's what you have but there are some
limitations that you need to recognize um you're likely only going to get aqueous metabolites and we have to
use a little bit of a different type of pulse sequence on the NMR to suppress some of those large molecule
signals and that does limit our ability to get kind of quantitative data and detect some of the lower abundant
samples and this is an example of a spectrum from a nun extracted sample and and you can see some of the
aqueous metabolites like lactating glucose but you can also see some of the lipid signal and you see kind of
this large kind of mounds that the baseline is is is raised and that makes quantitation of metabolites very
difficult so you can certainly get metabolite data but you're likely not going to get quantitative metabolite data
from this type of sample at least by NMR and you won't get a new specific lipid information and as I mentioned
you're you're going to have difficulty quantity quantifying so because of that we like to extract our samples and
that means that we want to precipitate out those large molecules that tend to obstruct a large portion of the
NMR spectrum and we use a methanol chloroform extraction to do that and what that allows for is it generates
an aqueous phase sample and a lipid base sample so we actually generate two fractions from a single sample
and then we add an internal standard and then we that allows us to align the spectrum and quantify the spectra
using a quantitative commercially available software platform and we are using kinoma key and the Cano miix
library is over 300 metabolites so we reference every channel thing to their existing library but this allows us to
generate named and quantified metabolites for each spectrum that's generated from your sample but notably
this is for the aqueous phase of your sample so this is this is what happens you have a blood or plasma or
serum sample you do this extraction and you generate a lipid and out and a water phase and so this does have
some potential advantages one it allows you potentially to run that lipid phase on another assay you can send
it to the MRC two for either targeted or untargeted lipidomics SI you can run your aqueous phase by NMR and
because NMR is not destructive you could also then subject that sample to LCMS either by a targeted analysis
or a complementary untargeted analysis so this allows some flexibility and how you use your sample there I
just want to mention that there are some disadvantages to the extraction even though you have the advantage
of generating these two phases the extraction itself is relatively time-consuming it's a little over 24 to 48 hours
depending on the humidity because the samples have to drive so like in the summer sometimes it takes a little
longer to get those samples dried and we we acknowledge that we probably are losing some metabolites we
don't know what for sure but likely some metabolites bind to those macromolecules that we pull out of of a
sample so the workflow for NMR metabolomics is very similar if not identical to that of lc-ms you you get your
sample you store it is very same way as you would for lcms we do an extraction just like for lcms there's a data
collection in this case NMR is the assay and then there's the data processing and analysis and I'm not going to
spend a lot of time on this because you've heard a lot about this over the last few days but this is just a
representative NMR spectrum and there's a number of things that we can do with the spectrum we can
quantify HP and metabolites and generate data that way and that that uses the kinome Excel for that I
described previously or we can do something like a binning approach using a chemometric or PCA analysis
and this can be done if you particularly if you have a large number of samples to quantify a large number of
spectra can be quite time-consuming so an alternative approach would be is to use this process called binning
which sets specific bin sizes to a specific range of ppm so for example point zero four ppm and the software
will then bend the spectrum and generate a area for each bin and a bin number and you can use those actually
as your data you can subject those data to chemometric analysis and from that analysis you can determine
which bins might be separating there's your groups and this is just a example acute lung injury versus sepsis
versus healthy patients the PCA would generate the list of the tablets it would be separating those samples or
the bins in this case for Benny and then you could go back to those bins and quantify the metabolites and
those specific bins so that can often result into just a more efficient processing of your samples so I wanted to
talk about some examples of work that we're doing where we're really using the complementary approach of
NMR and lcms and i'm working with a vascular surgery group in age-related deep vein thrombosis and for this
study we use serum samples that were collected as part of a biomarker discovery study so these were
somewhat samples of convenience and the there's a clinical problem or question in which there's patients who
are younger that developed DVT and older patients who develop DVT and it's not really clear what the
pathological underpinnings of those differences are they they tend to have different prognosis and there's no
definitive biomarkers for differentiating those mechanisms so in a very preliminary study we assayed some
samples from what we categorized as young DVT patients and old DVD patients and we extracted them and
we identified and named 38 metabolites using the kinome except where platform and then we did statistical
comparison of the two groups and we identified three potentially for metabolites that might differentiate these
groups and these might be familiar to you if you recently read the nature paper on cardiovascular disease and
carnitine and the meat diet and the microbiome but these were the ones that lit up in this small group of
patients carnitine was elevated in older patients so is calling and betaine and TMAO which we cannot
differentiate by NMR because those Peaks overlap so we have to name it as it as a dual metabolite we sent
technical replicates to the MRC to LCMS and working with Steve Brown and an lcms targeted assay that they
had developed and we confirmed our findings these are the quantified data by lc-ms and essentially
corroborated our NMR data so I think this is a really nice example of how we just did a preliminary study by
NMR found these metabolites and then we're able to use technical replicates illustrating the points of having
technical replicates and sent it to lcms to confirm what we had found by NMR Charles talked to yesterday
about work in a RTS and we are also working on a project i'm using serum samples and for this work we're
focusing primarily on lipids and NMR doesn't have a good history in measuring lipids and in fact it's somewhat
of a new application so we're working on developing some new techniques by NMR that we hope might
actually help inform a more targeted lc-ms lipidomics approach and for this study these were samples that
were collected in a larger study that was conducted at the University of Mississippi some patients with sepsis
and some patients with sepsis induced a rds.a RDS acute respiratory distress syndrome is an unfortunate
complication in some substance patients which for which we presently have no biomarkers whatsoever or no
no therapy so we don't really have a good therapeutic target for these patients extracted these samples as well
and we ran both the aqueous and lipid phases by NMR in this study and we found no difference in the aqueous
metabolites but there were some differences in the lipids and these are shown here phosphatidyl serine and
some the fatty acids signal by NMR were different and these are we had two time points an early time point
which we categorized as x 0 these were patients who presented to the emergency department so they were
fairly early in their course of illness and 72 hours later while they're in the ICU and we saw an early
differentiation in both these measurements in a RDS and substance patients that increased as time went on
and what we're doing now is we're using technical replicates to more fully elaborate this data set using the
lipidomics platform at the MRC to that I know someone talked about now I can't remember who it was sorry I
believe yesterday who was it that tuck just ah yes we're using that platform TM did that's right as an extension
of some of this work as i mentioned we're developing some NMR techniques to see if we can better inform
lcms lipidomics and there's some pulse sequences that we can use on the NMR to get better resolution of
lipids in the lipid signal and one of these is called diffusion ordered spectroscopy and it hopefully we haven't
we're just beginning to try it but our question is can this help us better separate some of these lipid signals so
we can more confidently identify specific species of lipids and it uses diffusion coefficient data to help
differentiate large versus small metabolites so we're going to give it a try and this is just a representative kind of
concept picture of what this might look like this is a standard spectrum and you can see there's there's a fair
amount of noise this is another type of pulse sequence that tries to reduce some of this uneven baseline we
get a little bit better better resolution and then you can see how a diffusion edited or a diffusion spectroscopy
approach might actually increase the resolution of some of the lipid Peaks so these are some of the things that
we're working on and trying to see if we can get more defined lipid data to help inform a more targeted lc-ms
lipid approach the other thing we can do by NMR is typically we run a one dimensional experiment we can also
run a two dimensional experiment which gives more detailed information and generates a more detailed
spectrum about specific metabolites and we can also run a p31 experiment to identify phospholipids so these
are some of the other techniques that we can use the experiments are longer a little more time consuming but
we typically get more and rich information about specific species so we're trying some of these types of tools
as well and lastly I wanted to talk about another project in sepsis pharmaco metabolomics you may be familiar
with pharmacogenomics we're taking the pharma part to the metabolomic platform and in this study we're using
complementary NMR and lcms assays to try to get more at drug action and in particular try to differentiate drug
response to our emmett goal is to bring a precision or personalized medicine approach to sepsis which
presently there is we have no pharmacotherapy really for these these patients so this project involves the use
of testing of a drug called l-carnitine or a nutraceutical l-carnitine as you may know it is required for the
transport of long-chain fatty acids into the mitochondria and it may it may be in novel therapeutic and sepsis
because there's some data to suggest that in sepsis there's a disruption of mitochondrial function that might be
rescue a bull by carnitine and so this could be not only a novel therapeutic but would be a much needed
therapeutic because half a million people in the United States die each year from sepsis and we really do not
have any effective pharmacotherapy so in this study we assayed the samples that were generated from the
phase 1 clinical trial which was a small trial of 30 patients 16 patients treated with l-carnitine and 14 patients
treated with placebo and in this study there was a modest reduction in mortality and the l-carnitine treated
patients we initially found just by comparing l-carnitine to placebo that there were there were some changes
you might expect in psalm metabolites but it wasn't really dramatic so we went back and we looked at survivors
who received l-carnitine versus the non survivors who received l-carnitine and we found several metabolites
that were different and these are times zero so this is pretreatment metabolite data from those who survived
and got l-carnitine and those that did not three hydroxy butyrate which is a one of the ketone bodies and it's to
do acetate which is another ketone body or different in survivors and nonsurvivors and we also found that three
hydroxy I so valerate was different and creatine was different and notably the acetyl carnitine carnitine ratio
which is believed to be a marker of carnage homeostasis was higher in the non survivor patients suggesting
that they may have greater difficulty or inability to rid themselves of acetyl carnitine so what we did from that
point is we went back and we looked at all the time zero metabolites we loaded that into metabolized all I
talked about metabo analysts yesterday and we look to see based on that time zero profile what pathways
might be the most significant and what we might focus on as we move forward and so this is the topology
network map generated by metabolize is an impact factor on the x-axis and a negative log P on the why and
the most impactful and significant pathway was the synthesis and degradation of ketone bodies so what we did
then is we went back and we categorized all the patients regardless of their outcome based on their time 03
hydroxy butyrate level and we found that in patients who had this low ketone low three hydroxy butyrate and
where carnitine treated had about a twenty percent mortality rate compared to both low and high ketone
patients that either got carnitine or placebo had a roughly eighty percent between seventy and eighty percent
mortality rate so this is just an example of how the metabolites in form of drug response that you may not
necessarily see or be able to detect at time zero with some more traditional markers because in fact other
clinical parameters like lactate levels which are often used to differentiate substance patients and sequential
organ failure assessment scores or not different in these patients at time zero so possibly the metabolites have
some information that could inform us of the likelihood of patients who may respond or be more responsive to
l-carnitine as an extension of this work we're hoping to go on to sa the samples from the ongoing phase 2 trial
which is a much larger 300 patient study and get some additional information as part of this we plan to do an
untargeted LCMS approach in collaboration with dr. Evans to elaborate this data set and also maybe do some
targeted LCMS essays of carnitine metabolites to more fully elucidate the potential mechanisms that might
underlie these differences between these these responders and nonresponders of l-carnitine so in summary
proton which is the most common NMR that's used for metabolomic can be used as a standalone certainly but
i think the strength and kind of potential of NMR isn't in concert in complementary approaches with LC ms to
generate a more elaborate data set of metabolites and i think that that analytical capability is advancing and
evolving and we're learning about different ways and new ways to apply NMR two different types of samples as
well as to generate different types of datasets particularly in lipidomics and if you're interested in lip in NMR
metabolomic sime here and this is my email address and lisa is here as well and please feel free to contact us
we'd be happy to talk to you and give you some advice or work with you however you'd like and I have I can't
leave without acknowledging a lot of people a lot of people who work in my lab very hard on generating these
data the support I've received and the numerous collaborations both here at the University of Michigan as well
as other institutions make this all possible thank you very much for your attention and I'd be happy to answer
any question yes yes yes great right it's an excellent question and I don't think Steve Brown is here because he
would know the answer that question we've been we've been doing these experiments where I will run a
sample in the College of Pharmacy NMR and then I'll take it over to the MRC too and we don't have those data
yet but I expect to shortly we're doing these types of kind of validation experiments to get an idea of how stable
the sample is because once we extract it this is an extracted sample the deed is done there's true we don't
think there's a lot in there that could really happen but we don't know for sure so it's not like we're taking a a
metabolically active sample and assing it by NMR and then you know I take you know 45 minutes to walk over
to brown right so we expect that we would be able to use that type of strategy run it in NMR and take that exact
sample and run it by lc-ms and I would I know you there's a lot more lcms expertise in this room then and I
have so yeah yeah and you know anything that's volatile or I think it's gone you know but it could be helpful you
know because what you're going to get from lcms that you can't get with NMR is you're going to get greater
sensitivity so what we pick it kind of like that first tier by NMR and then accepting some loss along the way by
lc-ms you're just you're going to be able to elaborate that so particularly if you don't have a technical replicate I
think the ideal situation is to have technical replicates because that really would be the best way but if it's not
possible or you don't have them I think it could be a viable alternative highly reproducible we do not see a lot of
variability NM ours of course service yes it's it's it's it's maybe a little cheaper but you know it's I think roughly
about a hundred dollars a sample it depends on your sample though and whether we have to extract it or not
so we can we can talk any other question yeah yes autosampler yeah yeah our Auto samplers for NMR we
have a small autosampler we can run 12 samples if you're at all familiar with the University of Alberta's NMR
center so the University Alberta has is Canada's NMR core lab essentially Nanook and they they have it's like
Houston and it's like a Space Center and they have a phenomenal arm set up with with an autosampler and
they can run Sam you know samples overnight and multiple samples for for for a 1d proton single sample that
it's not a relatively to ten minute run it's not a long run if you're doing 2d experiments and more elaborate
experiments those runs are going to be longer so that's when it can get a little more tedious with kind of
samples going in and out of the NMR yes I'm right right over there college of pharmacy yeah on the on the 500
we we have a cent you know there's an autosampler but there's always someone in the NMR lab so we can we
can do it one by one it's not it's not horrible so for for cell culture um the the volume the sample volume five
hundred microliters is best because we use a five millimeter tube and a five millimeter probe which we think is
the best way to do metabolomic saan NMR the three millimeter tubes and three millimeter probe it can be done
but we don't think the data are as reliable is off a five millimeter probe so that tube in order to to get the sample
you know where it needs to be five hundred microliters is ideal oh how many cells okay so for a plate we use
the larger plates I have to look at our protocol I don't know exactly what the cell numbers are but they're they're
similar to what I think you guys have been doing for lcms we have we have program we have SOPs for blood
collection if you guys you know if you want any of that we have all that and we're working on cell culture model
we're instead of using organic and extraction processes or using sonication strategy to generate an aqueous
lipid and media phase from cells and c-13 glucose studies and we're going to run those samples friday and just
see what we get we don't know we're going to get yet yeah yeah I got to do that you're really good um but you
know so if depending on what Reed we get on NMR we probably like to run those samples by lc-ms just to see
we get even a better signal because I honestly don't know we might not get anything by NMR I don't know we'll
see yeah we'll run a seat will right run a proton on a c-13 yep yep any other question thank you