CyTOF XT Performance in comparison to Helios

[Jahangir]

Hi everyone,

As the CyTOF XT has been around for a couple months, I'm just starting a thread so that we can all share our experiences with the CyTOF XT, how it compares with the Helios, if you've had any troubles with the new XT and how you've troubleshooted those problems, what's easier and worse at a user/operator level than the helios, etc.

As for myself as a user, running complex epithelial cell samples and with co-cultures with fibroblasts, the XT is much more sensitive to becoming blocked (esp the nebuliser) than the Helios+Supersampler combo. The clearing clog procedure is not used at all - we've found it just causes the nebuliser to become blocked (with our samples at least). So we always stop this function and try to clear it ourselves using bleach or removing the nebuliser and manually removing the block - which works (most of the time). We've realised that we need to keep an eye on the XT much more than with the Helios+Supersampler combo, ensuring the pressure doesn't increase too much.

We've also observed that the XT calibrates/tunes the 190+ mass range differently to the helios. My reason for saying this is because we use the Cisplatin range available to us (i.e. the 194-/195-/196- & 198-Pt) as a live/dead cell stain but also for barcoding (I make the TOBis barcodes myself in-house). We also use the 191-Ir and 193-Ir channels for DNA staining. We've observed that the absolute intensities for all the stains I've mentioned have decreased by literally one order of magnitude. So the iridium intensities used to be 10^4 with the helios, but now it's at 10^3. The cisplatin intensities for our barcodes used to be between 10^3 and 10^4 but now we're only getting an intensity of 10^2 - this has caused the debarcoding efficiency to decrease slightly as well. Does anyone why this could be? I'm currently in the process of QCing and adjusting the concentrations of cisplatin in my barcodes to ensure that the dynamic range between the positive and negative signals are more distinguishable as a resolution to this problem but it would be nice to know why there's a difference in the 190+ mass range between the XT and helios.

However, not all bad news. The carousel works amazingly, and I love how (nearly) everything can be done remotely - including washing and shutting down the XT.

It would be nice to hear the experiences from other users and also if they've observed the same differences in the 190+ mass range.

Best regards,

Jahangir Sufi

[mleipold]

Hi Jahangir,

Out of curiosity, are your FCS files normalized during output (ie, ending in "_Processed")?

The CyTOF XT software Help (FLDM-00045 Rev 02) p. 93 says "When selected, an FCS file is created for each sample in a batch. Events, Randomize, and Normalize are selected by default, creating a single FCS files with all three processing functions applied. This file name is appended with _Processed." In this case, the Normalize is done by the EQ6 beads ("The default bead type for CyTOF XT™ analysis is EQ6 beads. EQ4 beads are added to samples prepared with a Maxpar Direct Immune Profiling Assay kit.")

If so, have you compared your Raw (not _Processed) and seen whether this decrease is still observed? I'd be surprised if a full log decrease would be caused by Fluidigm-type Normalization (now that 209Bi is included in the EQ6 and can better norm that region), but it's something easy to check.


Mike

[LukaTandaric]

Dear Jahangir and Mike,

for some added data from our XT about the situation Mike is describing, attached is a comparison of the 190-209 channel range data between the [Helios], [XT - Pre-Processed] and [XT - Raw] I did on a sample I split directly before acquisition to acquire half on the Helios and half on the XT.

The data is of cleaned-up (event length, beads, gaussian, iridium doublets and debris) singlets from a healthy donor (HD) whole blood sample.
The only channels I actively had staining on were 191Ir, 193Ir and 209Bi. The lead I will address later.
I am not seeing a signal intensity difference between the Helios and XT specific to the 190-209 range.

I will make another reply soon with much more data about this comparison I did, because I used my full panel (89Y - 209Bi, including cadmiums, but not including platinums). I found there to be some notable differences between our Helios and XT.

- Luka Tandaric, University of Bergen, Norway

[LukaTandaric]

As per my previous reply, to compare the quality of the data generated from my 40-marker panel between the Helios and the XT, I had fully processed a single larger sample consisting of three different samples barcoded together (buffy coat, bone marrow, PHA-stimulated PBMCs), and split that sample into two halves only at the point where the sample is placed into the machine.

I have observed the following (table and figures in attachment):

A) The median signal intensity for all channels taken together was 25-30% lower on the XT, compared to the Helios (table 1). I attribute this to the fact that, for this particular experiment, the Helios had tuned well, compared to the XT, which tuned not-so-well (fig. 1, fig. 2). A repeat of this kind of experiment was done very recently at our core facility, where the opposite was true - the median signal on the XT was slightly stronger than on the Helios; again, due to the difference in tuning results.

B) Visible in the XT data, there was a large amount of contamination from 138Ba, 139La, 140Ce and 208Pb, which was not present in the Helios data (fig. 3). A peculiar thing is that this contamination is not present as a constant stream of ions in the rain plot, but only shows up on cells (fig. 4, fig. 5). The spillover of the listed contaminants causes a constant and strong false-positive signal in their spillover channels, most notably in 142Nd (from impurity spillover of 140Ce to 142Ce), and 155Gd and 156Gd (from oxidation spillover of 139La and 140Ce, respectively) (fig. 6).

C) Compared to the Helios, which produces nice, even data when viewed on a biaxial plot with time as one of the axes, the data from the XT takes the form of what I can only call bubbles. As I understand, this is due to the fact that the XT injects the sample in smaller aliquots, unlike the Helios, which introduces a constant stream of sample into the apparatus. (fig. 6)

D) For the offset gaussian gate, the Helios seems to follow the appropriate formula (offset = 100 * [absolute value of the vertical offset of the base of the event's signal distribution from the general baseline]). However, this either does not hold true for XT data, or the XT produces data with a very small vertical offset of an event's signal distribution base between 0 and 0,2. All other gaussian parameters are much less different between the two machines. (fig. 7)

I kindly appreciate any insight and speculation into these observations.

Kind regards.
Luka Tandaric, University of Bergen, Norway

[mleipold]

Hi Luka,

Thanks for all the analysis! A few initial comments and followup questions:

1. Regarding the backgrounds in 138/139/140/208 in point B: you said that the sample was processed, and then split immediately prior to acquisition on each instrument. Just to clarify: what was the sample resuspended in? MillIQ? CAS? CAS+? Did that differ between the two instruments, in case that's the source of the contamination (rather than the instrument itself somehow)? But I completely agree that the XT data has the issue while the Helios data does not.

2. Regarding the "bubbles" in point C: yes, presumably the "bubbles" with some minor gapping are the result of the XT using a sample loop that refills, vs the continuous sample delivery of the Helios PSI.

3. That said, I'm a little surprised at the inconsistency of the 138/139/140/208 background vs the markers such as 209Bi: for example, in Fig 6, the 209Bi signal is pretty steady near 1e2, whereas the 142/155/156 background channels all have a dip at the beginning of the loop injection. This is also seen in Fig 3, especially in the 139 and 140 channels.
- I don't have a good explanation of what's going on here: I agree that the background signals appear to be cell-associated rather than "streaking" as shown in the Rain Plots (Fig 4 and Fig 5), but why the cell-associated Background would have this "dip" and the cell-associated Marker signal *doesn't*, I have no explanation.


Mike

[mds4z]

All

This is very helpful information as we are getting ready to trade in our Helios for an XT. Any additional feedback or comments would be greatly appreciated before we make this decision. Reading this we are somewhat reluctant to make purchase. All the XT automation is a key feature, our hope as a core facility we will be able to allow well trained users operate without core staff present, reducing operational costs for our users. Is this wishful thinking?

Thanks
Mike

[cer525]

HI Mike S.,

In case you have not come across this yet on the Standard BioTools website, there are 3 recordings from current owners of the XT that each have sections describing performance and how the instrument is being used in their lab or Core. You can find them here: https://www.fluidigm.com/products-servi ... ars-anchor

I hope this helps,
Clare Rogers
Standard BioTools

[silatour]

Hi all,

We got a Cytof XT few month ago and so far I'm not happy at all about it. I don't trust the XT yet to run precious samples (or more generally any samples).

The main problem we observe is clogging. We observe clogging in almost every samples we run where we never had problem with the helios.
I think most of the clogging happen in the nebulizer and it look like the unclogging protocol doesn't resolve these clogging.

If any of you have any tips on how to avoid that it will be awesome as so far I prefer to run on the helios and spend hours in front of the machine instead of using the XT and its autosampler.

Cheers

Simon

[mleipold]

Hi Simon,

To better understand: what kinds of samples are you typically running?

PBMCs or other regular suspensions? Or disaggregated tissue?


Page 27 of the XT USer Guide (FLDM-00254) says the following:

"The Autosampler begins the automated clog sensing workflow by switching from injection mode to load mode, which enables the push pump to be used to flush the nebulizer while the sample in the sample loop remains undisturbed. An initial attempt to remove the clog is done by pulse aspirating and flushing at high flow for about 1–2 min. If the clog is not removed, the software then runs a clog identification workflow, in which the 2 pressure sensors are used to determine whether the clog is in the sample loop or downstream of the sample loop (for example, in the nebulizer line or the nebulizer). The unclogging procedure then focuses on the location of the clog. During the unclogging, the system increases the flow rate level and the pulse rate from light to more vigorous."

However, it's not clear to me *at all* how you can flush the nebulizer by "pulse aspirating and flushing at high flow for about 1-2 min" at a "pulse rate from light to more vigorous" can be done without blowing out the plasma.......I thought the anti-clogging mechanism only operated "inside" (ie, prior to the "grounding nut" type of area).


Mike

[silatour]

Hey all,

We run mainly cell lines (293T cells, THP1 and cancer cell lines)

Fluidigm engineers/Field Application Specialists worked in close collaboration with us to identify the problem and we identified the origin of the clogs.
Basically after IR, we were washing with PBS BSA at a too high concentration and then using PBS instead of the recommended CAS+ and that lead to a lot of clogging.

So don’t try to go cheap using PBS, definitely go for the CAS+ solution as that is a game changer regarding clogging.

Since we modified that, I was able to run multiple batches of 16 samples without any clogs (or small clogs that the XT resolved himself). I just load the samples in the carousel and run away and the XT run overnight. Last week I acquired samples for 12H in a row without any problem.

For the comments about contamination in the post above, we don’t observe anything like that in our machine, the contamination level are low (never looked at the actual values, but definitely not as high as what is show in figure 3).

I have to admit that I totally changed my mind since the clogging issues got resolved thanks to Fluidigm peoples.