High-Throughput Phospho-Flow Mass Cytometry for Immune System Mapping

Automation and Standardization

Zach Bjornson

the goal

To survey the differences in cellular immune behavior between species used for therapeutic development and humans, as well as within populations.

the animal rule

how good are the models?

  • 2003 to 2011 drug approval rate (P1 to IND): 9%
  • Models might appear similar to humans structurally at a high level, but many known cases in which they differ.
  • If 91% of drugs work in models, but not in humans, what percent would work in humans, but not in models?

the approach

Stimulate blood from humans and animals with a panel of immune stimuli, then use phospho-flow assays read out by single-cell mass spectrometry (CyTOF).

challenge:

Consistently and quickly assay
5,248+ samples

protocol, in brief:

  1. mix blood with stimuli
  2. incubate for 15 minutes at 37 degrees
  3. add formaldehyde to fix cells
  4. lyse red blood cells
  5. wash twice
  6. barcode cells (multiplex)
  7. wash twice
  8. combine samples
  9. add blocking reagent
  10. add extracellular antibodies
  11. incubate
  12. wash once
  13. add methanol to permeabilize membranes
  14. wash four times
  15. add intracellular antibodies
  16. incubate
  17. wash once
  18. add DNA intercalator

stimulation variability

Task: Dispense and mix different stimuli into 96 samples, incubate at 37 degrees for the same amount of time, then fix after exactly 15 minutes.

Same lot of each stimuli and diluent shall be used.

Using a 12-channel pipettor for 8 rows takes about two minutes—too long.

Solution: Single-use "cartridges" loaded with stimuli and diluent administered by 96-channel pipetting.

reagent variability

Task: Expose all 5,248 samples to the same antibodies at the same concentrations.

Reagents evaporate over time (1-2% per month at 4 degrees).

Large antibody cocktails have questionable stability.

Human assembly of cocktails is subject to errors and pipetting variability (~1.5%).

Solution: Lyophilized cocktails and sample multiplexing.

the robots!

layout planning

remote monitoring and error paging, a custom VWorks plugin

instrument running time

Task: Acquire 5 million cells per donor for 328+ donors.

Typical CyTOF acquisition rate is 1 million events per hour
(68 continuous days or 164 10-hour days).

Human must inject sample every 10 minutes.

Cell flow rate is highly variable over the course of every 10-minute injection.

Solution: New fluidics system design.

The "Super Sampler"

The "Super Sampler"

Hardware and software that continuously runs samples at the optimal rate.

89% improvment in throughput on CyTOF 1, 68% on CyTOF2 (664 hours of run time saved)

Unattended operation for hours at a time

Software pages a human if a problem is detected.

After all that, how is the consistency?

No detectable run-to-run or well-to-well batch effects