Quantifying Metal Mass on EQ Beads

[ndonahue]

I am trying to obtain the masses for Ce140, Eu153, Ho165, and Lu175 on the EQ beads. I ran dissolved metal standards for these elements at known concentrations to establish a calibration curve in solution mode. I then ran the EQ beads in event mode. I hypothesized that if I have known ion counts from the standards I can obtain the mass of metals on the beads.

With the data I have now, I can calculate the TE for the dissolved standards. From there I am kind of lost. I am following the equations/rationale in this paper where they quantified silver nanoparticles per cell. https://pubs.acs.org/doi/10.1021/acs.analchem.7b01006.

Another study I am trying to follow is here: https://pubs.acs.org/doi/10.1021/ja9052009 where they calculated the number of metal ions on microspheres with CyTOF.
Any thoughts or insight is greatly appreciated! Thank you

[mleipold]

Hi Nathan,

What is your ultimate goal for this experiment? To characterize your instrument? To characterize EQ beads? To use EQ beads as an internal quantification reference for another experiment (an alternative would be metal-labeled antibodies on capture beads with defined number of capture epitopes)?

A few other comments:
1. You have to be careful about the linearity of the instrument. In the Ivask et al paper, Fig 1b shows the cells have an Ag signal above 1e4 "Ag Counts" (I assume this is Ag Dual Counts). The CyTOF community knows from Iridium and some other measurements that the linear range of the instrument tops out somewhere in the 5000 Dual (ie, 0.5e4, or 2-4x lower than in the figure). By this, I mean that if you overstain your Ir, you'll start to see one of the Ir isotopes go "off diagonal" as its detection saturates. So, your quantitation will not be as tight as you (seem to) want in that area.

2. I'm not sure it's as straightforward to convert from Solution (metal solution standards) to Event (beads) as described in Ivask. Solution mode takes a defined window of time and quantifies whatever is inside that time window, while Event mode has some additional definitions (EL 10-150, LCT). Additionally, there's a distribution in sizes of the EQ beads, including metal content (the peak is narrow, but not infinitely so and not to the level of a Metal Standard from NIST). Therefore, you're getting a range, so you'll have to use Medians and such.
- Whether this range is sufficient to make your calculations only semi-quantitative, I don't know.

3. You might take a step back and try this with antibody capture beads and an antibody-metal conjugate. That way, you can meaasure:
a) a metal solution standard - Solution mode
b) the antibody stock used to make the capture bead sample - Solution mode (additional source of error: how well you can define the antibody protein concentration)
c) the capture bead sample - Event Mode

4. I'm not sure what you mean by "analog-to-digital intensity factor"....do you mean Dual coefficient, for converting from Pulses to Intensity?


Mike

[ndonahue]

Hi Mike,

Thank you for your great reply! I really appreciate it. The objective for running beads + metal standards on CyTOF is to validate another quantitative multi element technique for single cell analysis. CyTOF is the established gold standard for single cell multi element measurements and I was hoping to obtain similar values between two different techniques.

1. The highest ionic metal standard we ran was 1 ppb. The lowest standard was 0.1 ppb which was above the background. Fortunately, we avoided saturation with these dilute standards and stayed within the linear range.

2. I agree it is definitely not straightforward to convert values from solution mode to event mode and vice versa! I ran the EQ beads in solution mode & event mode (at a concentration of 100,000 beads/mL) for 2 minutes and can see signal variation. A range/distribution of masses is fine by me and was expected based on my previous data where I ran EQ beads on an ICP-MS in single particle mode. I am expecting ~ 2000-5000 attograms of metal per bead, but my CyTOF calculations using the aforementioned papers fall short by an order of magnitude.

3. I like the idea of using the capture beads and a metal conjugated antibody. This could work on both CyTOF and the other instrument I am trying to validate.

4. Yes the "analog-to-digital intensity factor" from the 2009 JACS paper is almost certainly the dual coefficient. In that paper they have the equation:
N= (I x IF)/T
where N is the number of metal ions per bead, I is the mean intensity from the TOF detector (sounds like solution mode to me?), IF is the "analog to digital intensity factor" (probably dual coefficient nowadays), and T is the transmission efficiency (obtained from metal standards).

I was hoping that equation would be a panacea to this problem, but the IF term remains elusive to me.

Thanks again for your input!

Hi Nathan,

What is your ultimate goal for this experiment? To characterize your instrument? To characterize EQ beads? To use EQ beads as an internal quantification reference for another experiment (an alternative would be metal-labeled antibodies on capture beads with defined number of capture epitopes)?

A few other comments:
1. You have to be careful about the linearity of the instrument. In the Ivask et al paper, Fig 1b shows the cells have an Ag signal above 1e4 "Ag Counts" (I assume this is Ag Dual Counts). The CyTOF community knows from Iridium and some other measurements that the linear range of the instrument tops out somewhere in the 5000 Dual (ie, 0.5e4, or 2-4x lower than in the figure). By this, I mean that if you overstain your Ir, you'll start to see one of the Ir isotopes go "off diagonal" as its detection saturates. So, your quantitation will not be as tight as you (seem to) want in that area.

2. I'm not sure it's as straightforward to convert from Solution (metal solution standards) to Event (beads) as described in Ivask. Solution mode takes a defined window of time and quantifies whatever is inside that time window, while Event mode has some additional definitions (EL 10-150, LCT). Additionally, there's a distribution in sizes of the EQ beads, including metal content (the peak is narrow, but not infinitely so and not to the level of a Metal Standard from NIST). Therefore, you're getting a range, so you'll have to use Medians and such.
- Whether this range is sufficient to make your calculations only semi-quantitative, I don't know.

3. You might take a step back and try this with antibody capture beads and an antibody-metal conjugate. That way, you can meaasure:
a) a metal solution standard - Solution mode
b) the antibody stock used to make the capture bead sample - Solution mode (additional source of error: how well you can define the antibody protein concentration)
c) the capture bead sample - Event Mode

4. I'm not sure what you mean by "analog-to-digital intensity factor"....do you mean Dual coefficient, for converting from Pulses to Intensity?


Mike

[mleipold]

Hi Nathan,

For the Dual coefficient, could you use one of the ones from the Tuning QC report (pick the one that's closest in mass to what you're trying to measure)?

Regarding Solution vs Event: I wonder, if you set the Solution Mode acquisition time segment similar to that of a cell event (~20 pushes = ~260 usec = ~0.26 msec), would that give you better agreement?

I'm not sure how to do that in the Helios software, though: at least with the newer v7.0.8493, the only parameter that is given for you to change is when to start Dual.....even in Service mode.

With CyTOF1 and CyTOF2, there was a parameter called "Pushes Per Reading", but I think they got rid of (or hid) it.......


Mike

[AMitch]

Hi Nathan and Mike,

This is a bit of a late response to the thread. I’m a co-author on the Ivask paper.

First up, Mike’s comment about high Ag signal being beyond the limits of linearity - this is a valid criticism. At the time the experiments were done I wasn’t clear on the limits of the linear range. It looks like we chose the highest value for illustrative purposes, and this will have led to underestimation of Ag signal for this condition, but the vast majority of data points should be well within the linear range. In any case I don’t believe this invalidates the generic approach.

In terms of absolute quantitation of signal from particles, in our hands the approach matches well with quoted/theoretical signal from test particles including gold nanoparticles of defined sizes, as well as EQ beads - definitely within an order of magnitude. For example, signal from 140Ce and 165Ho (metals with in-house available standard solutions to determine transport efficiency) were 64% and 103% of the values quoted on a typical EQ bead spec sheet. (note: unfortunately we don’t have EQ bead data matching specific batch specs, but assume that it wouldn’t vary wildly between batches... though this could be dangerous to assume). Feel free to direct message me to go through calculations if you want to.

Some other thoughts - happy to have others on the list comment if they agree/disagree:

* In either liquid or particulate samples, transport of metal atoms through the entire system is the same. This is definitely a simplification. Transport can be broken down into 2 broad steps - physical sample introduction into the plasma and ion transmission from the plasma to the detector. Signal from liquids is integrated over a long period of time (say 1 second) and this comes from a very large number of droplets generated by the nebulizer - only a percentage of which actually make it through to the plasma without getting lost on the spray chamber/injector - probably about 50-70%. In contrast, clouds of atoms from a particle should be largely discrete and either enter the plasma or not. This means the TE of fluids is likely to be underestimated by ~30-50% compared to particles. This isn't taken into account in the method in the paper.

* Does the presence of a particle in a droplet affect its transport to the plasma? In the single particle ICP-MS literature, the efficiency of sample transport to the plasma is generally estimated based on the number of test particles - typically EQ beads for mass cytometry - detected, vs the number entering the nebulizer (based on sample flow rate and known concentration). If, for example, having a particle in a droplet leads to it being more likely to hit the spray chamber wall, then liquid and particulate sample transport efficiencies wouldn’t match. There’s an indication in the single particle ICP-MS literature that this might be the case under some circumstances, but hasn’t been looked at with the sample introduction system of the CyTOF/Helios.

* Is the transmission of ions from from the plasma through to the detector the same for particles and solutions? This is probably true, but it’s possible that a large cloud of ions with a short time duration that derives from a particle might behave differently to a low level continuous signal from a solution. For example, transport through the pinhole in the sampler cone might be less efficient for a cloud.

* In practical terms, for the particle detection approach used by Flulidigm, thresholding parameters are important but probably don’t greatly impact quantitation of cell associated signal under typical settings/conditions. The setting with the most impact is the lower convolution threshold. Based on discussions with our local FAS, event detection is initiated after the push number criteria for the event is satisfied, and when the signal reaches the LCT threshold. Collection is continued until it falls below the LCT again. Signals from all channels between these two points are recorded in the event. This matches in-house testing using defined particles where increasing the LCT leads to decreased event length and a drop in per-particle signal as the tails of the pulses are cut off. Alternatively, keeping the LCT the same but increasing the signal contributing to thresholding (for example, by adding channels one-by-one using custom expression filtering in imd processing) also increases event length/integration time. However, the potential errors introduced by thresholding seem to be no more than 10-20% even if they’re much higher than would be typically used.

Andrew

[mleipold]

Hi Andrew,

Thanks for the great reply (and clarification)! A couple further comments.

1. Re EQ bead lot: to the best of my knowledge, there has been only one batch synthesis of Fluidigm EQ beads ("bucket"). This is why so far there's only been one Passport for normalization.

From what I've been told, the different Lot numbers are different "dips" out of that same "bucket"....while the metal content of the beads in a neutral solution should be pretty stable even over years (as demonstrated in some of the early Abdelrahman and Thickett papers from the Winnik lab), formally it would still be something Fluidigm needed to track.

However, assuming Fluidigm *does* release new EQ beads containing Y, In, and Bi (see viewtopic.php?f=10&t=1868 ), then I assume that a new Passport(_ver3?) would be required: I would assume a slight variation even in the Ce/Eu/Ho/Lu content between the two types of beads, since nothing is ever *perfectly* reproducible.


I assume that Fluidigm will correct my statements if they are wrong.


2. Regarding sample+ion transmission efficiency, Solution vs Beads: I would think that the sample transmission efficiency of the droplets (Solution *or* Bead-containing) from the nebulizer to the plasma (nebulizer, spray chamber, injector) should be the same.

I do agree that the efficiency of a continuous line of ions from a Solution may differ from the efficiency of discrete clouds/bundles of ions from Particles. I'm not sure which to expect to be higher efficiency getting into the Sample orifice: Solution would be relatively unchanging, while Particle Cloud efficiency would depend partly on total ion cloud expansion related to total metal content (as well as specific path/transit time through the plasma, one reason why the tighter focusing of the HT injector changed the Event Length).



Mike

[AMitch]

Hi Mike (and Nathan),

That’s good to know about the EQ beads. The spec sheet that I have is from 2013 so that’s so I’d assumed that they’d made another batch in the meantime!

In terms of the original question in the thread - in the spec sheet the quantification method that Fluidigm (or DVS as they were then) gives is “Beads have been independently characterized by ICP-MS (PE-ELAN-DRC) to determine the number of lanthanide ions per bead”. There isn’t a description on how this was done but it’s probably the method used in Abdelrahman et al. Journal of analytical atomic spectrometry 25 (2010): 260-268, which uses microwave digestion in concentrated HNO3. For what it’s worth, the spec values are listed as: 140Ce = 19.9E6, 151Eu = 11.3E6, 153Eu = 12.0E6, 165Ho = 7.6E6, 175Lu = 9.8E6, (+/-15% atoms/bead). Obviously bulk population measurements.

Andrew

[ndonahue]

Hi Mike and Andrew,

Thank you for the great discussion. The points you both have been bringing up are very informative. I hope my calculations from the Ivask paper are correct, but if it is possible Andrew thank you for being willing to check my math! After some more analysis, I was able to compare the mean values (ions per bead) between the CyTOF and single particle- ICP-MS. The graph is here https://imgur.com/zKN3bzo I'm not too satisfied with these results as the mean values between the two techinques on the same particles do not corroborate each other. I was thinking I could potentially plot the distribution of ions per bead, but am not sure of how that would work with the FlowJo files for CyTOF. It isn't a problem with single particle ICP-MS data.

Hi Mike (and Nathan),

That’s good to know about the EQ beads. The spec sheet that I have is from 2013 so that’s so I’d assumed that they’d made another batch in the meantime!

In terms of the original question in the thread - in the spec sheet the quantification method that Fluidigm (or DVS as they were then) gives is “Beads have been independently characterized by ICP-MS (PE-ELAN-DRC) to determine the number of lanthanide ions per bead”. There isn’t a description on how this was done but it’s probably the method used in Abdelrahman et al. Journal of analytical atomic spectrometry 25 (2010): 260-268, which uses microwave digestion in concentrated HNO3. For what it’s worth, the spec values are listed as: 140Ce = 19.9E6, 151Eu = 11.3E6, 153Eu = 12.0E6, 165Ho = 7.6E6, 175Lu = 9.8E6, (+/-15% atoms/bead). Obviously bulk population measurements.

Andrew