Protocol for separating cancer cell subpopulations by metabolic activity using flow cytometry

Summary Cells, even from the same line, can maintain heterogeneity in metabolic activity. Here, we present a protocol, adapted for fluorescence-activated cell sorting (FACS), that separates resuspended cells according to their metabolic rate. We describe steps for driving lactate efflux, which produces an alkaline transient proportional to fermentative rate. This pH signature, measured using pH-sensitive dyes, identifies cells with the highest metabolic rate. We then describe a fluorimetric assay of oxygen consumption and acid production to confirm the metabolic contrast between subpopulations. For complete details on the use and execution of this protocol, please refer to Blaszczak et al.1


Highlights
Metabolic activity is a key cellular phenotype, determined by a myriad of variables Cancer cells with higher lactic acid permeability tend to have higher metabolic rate

BEFORE YOU BEGIN
A key phenotype of cancer cells is their metabolic activity, which includes a description of lactic acid fermentation and mitochondrial respiration that supply energy and building blocks for proliferation.Metabolic activity is determined by the abundance, distribution, and activity of enzymes and transporters, as well as the spatio-temporal profile of substrates, intermediates, and products.This complexity provides scope for dynamic metabolic heterogeneity which manifests in cancers, ostensibly because it may offer a growth advantage.However, mechanisms of this heterogeneity are challenging to study because metabolic rate is difficult to resolve at single-cell level, let alone use it to separate sub-populations by metabolic phenotype.Single-cell transcriptomics and proteomics can describe the network of enzymes and transporters, but fluxes are not readily inferred from this information.Single-cell metabolomics provide a snapshot of metabolite abundance at steady-state, rather than their flux.
Attempts to separate cells by metabolic activity should consider a metric related to flux that can be implemented for sorting techniques, such as FACS.Fluorescent sensors of metabolite abundance are now available for real-time measurements, but sorting by these signals separates cells by steady-state metabolite levels, which is not necessarily proportional to flux.The challenge is that conventional flow cytometry takes a single measurement per cell, which -by definition -cannot interrogate changes over time.One strategy could involve a carefully timed experiment, where a substrate is added and its intracellular abundance probed flow-cytometrically after a fixed time delay.However, this may require cells to be substrate-depleted to assign a baseline: a maneuver that may affect metabolism.Moreover, many metabolites enter into a steady-state, from which flux is not possible to calculate without an intervention, such as enzyme inhibition.A concern is whether steps required for sorting introduce stress that compromises flux estimates from metabolite abundance, a relatively labile variable.
Recently, we presented evidence 1 that a cell's fermentative rate relates to the membrane's permeability to lactic acid (P Lactic ), a process facilitated by monocarboxylate transporters (MCTs).Net lactic acid production must equal its removal across the membrane, determined from the product of P lactic and the transmembrane driving force.A highly glycolytic cell benefits from higher P lactic because this prevents excessive accumulation of lactate and H + ions in cytoplasm.Indeed, hypoxic induction of MCT4 2 represents an effort to match lactic acid production with efflux capacity.Since P lactic is a property of the membrane, it is likely to be more stable than metabolite levels during flow cytometry protocols because internalization or trafficking are not likely to be significant during short protocols.Conveniently, lactic acid efflux across the surface membrane -irrespective of whether it is through the lipid bilayer as the undissociated acid or as H + -lactate co-transport by MCT -generates a pH change that can be measured using calibratable fluorescent pH dyes, such as cSNARF1, used widely for ratiometric fluorimetry.Indeed, the standard assay for P lactic is to measure the rate of intracellular pH (pHi) change triggered by a maneuver that alters the driving force, typically extracellular lactate. 3ased on these observations, we designed a protocol that approximates P lactic from the change in pHi in response to a carefully timed protocol that involves pre-equilibrating cells with lactate followed by rapid removal to drive lactic acid efflux.Cells that produce the largest alkaline transients have the highest P lactic .Subsequent metabolic phenotyping using our fluorimetric method 4 confirmed that these cells produce a higher fermentative rate, alongside higher respiratory rate, indicating a state of elevated metabolic activity.Strikingly, the metabolic contrast between sorted subpopulations was short-lived, which is consistent with dynamic behavior, whereby cells alternate between metabolic state.Our finding underscores the importance of sorting cells by a surrogate of flux, and presents a simple method of achieving high contrast between emergent sub-populations for subsequent studies.As an illustration of the utility of our method, we have been able to profile subpopulations for transcriptomics in order to interrogate the underlying mechanisms of differential metabolic activity. 1

Preparation of media and cells for sorting by metabolic activity
Timing: up to 1 week before sorting 1. Prepare ''lactate-loading medium'', ''lactate-free sorting medium'', and ''low-buffering medium'' by mixing ingredients listed in recipe in materials and equipment section.a. Once dissolved, heat solution to 37 C and titrate pH to 7.4 with 4 M NaOH or 5 N HCl, as necessary.b.Sterile filter using a 0.22 mm filter unit.
Pause point: Store medium at 4 C until use.
2. Seed cells at a density determined empirically to produce at least 18 million cells in 4-7 days.
a. Maintain cells in a standard culture medium appropriate for that line.
Note: Information regarding the standard culture medium appropriate for a given cell line should be obtained from the cell supplier.In this protocol, we use MIA PaCa-2 cells alongside a standard culture medium of RPMI-1640 with L-glutamine and NaHCO 3 -, supplemented with 10% FBS, 1% penicillin-streptomycin, and 1x sodium pyruvate.
b. Replace medium regularly to avoid excessive acidification by metabolic activity.
Note: At least five 15 cm dishes are recommended.
Note: High pH buffering capacity ensures pH stability that is necessary for performing calibration experiments.
a. Once dissolved, divide the solution between 8-10 beakers.b.Heat solutions, one by one, to 37 C and titrate to a desired pH with 4 M NaOH or 5 N HCl, as necessary.

Note:
The range for calibration should span from pH $5 to $9 at evenly spaced intervals.
CRITICAL: One calibration solution should be at pH 7.4 to represent physiological pH.
c. Record the precise value of pH attained to 3 decimal places.

CRITICAL:
The precise pH values will be used to fit the calibration curve.CRITICAL: Calibrations must be performed in the same type of plate as that used for metabolic phenotyping.Consider sterile, tissue culture-treated, black wall/clear bottom plates.
a.Under sterile conditions, thaw the HPTS/RuBPY stock mixture and dilute 1:1,000 (v/v) in the high-buffering media.
CRITICAL: Vortex the HPTS/RuBPY stock mixture thoroughly to ensure dyes are fully dissolved.This ensures that the molar ratio of dyes is preserved.
b. Load plate with high-buffering media containing the two fluorescent dyes.
Note: A 96-well plate and 100 mL solution per well are recommended, using at least triplicates per calibration point.
c. Place plate in a microplate reader with a dual gas controller at 37 C.The atmosphere should be CO 2 -free and produce regulated levels of O 2 .
Note: This ratio is not expected to change with O 2 partial pressure.
ii.In wells at pH 7.4, R O2 is defined as F 3 /F 4 .
Note: This ratio is not expected to change with pH.
g. Obtain calibration parameters by fitting HPTS and RuBPY calibration curves: i. Fit the relationship between recorded pH and calculated R pH to equation: where pK a , r max and r min are the calibration variables.
Note: Curve-fitting can be performed using a package such as MATLAB or bespoke methods.Exemplar calibration curves are provided elsewhere. 4.Normalize R O2 to recording at 21% O 2 and fit the O 2 -dependence with a line constrained to cross 21% O 2 at R O2 = 1: Extrapolation of the line to 0% O 2 estimates R O2 under anoxia (r anoxia ).
Note: Curve-fitting can be performed using a package such as MATLAB or bespoke methods.Exemplar calibration curves are provided elsewhere. 4A typical value for r anoxia is 0.7.
CRITICAL: Every set-up will have a unique calibration curve.Calibrations are not necessary after every experiment but should be performed routinely (e.g.several times a year) or after major changes in equipment, including upgrades and servicing.Note: Volumes of 0, 50, 100 and 150 mL are recommended.
CRITICAL: Dispense the oil gently by touching pipette tip against well wall.
c. Deplete O 2 in the medium beneath the oil barrier by equilibrating the plate for at least 6 h at 37 C under a CO 2 -free atmosphere with minimal or no O 2 .

Note:
The incubator must have regulated O 2 levels.
d. Remove plate lid and rapidly transfer the plate to the microplate reader with dual gas controller at 37 C and maintain a CO 2 -free atmosphere with 21% O 2 .This drives O 2 ingress down a partial pressure gradient across the oil barrier to the hypoxic medium.
CRITICAL: O 2 ingress will begin during plate transfer, therefore this step must be as fast as practical.
e. Measure fluorescence at regular (e.g., 1-5 min) intervals, until a plateau is attained: i. 510 nm emission excited at 416 nm (F3): pH-insensitive HPTS fluorescence; ii.620 nm emission excited at 450 nm (F 4 ), O 2 -sensitive RuBPY fluorescence.f.Export measurements and calculate ratio R O2 as F 3 /F 4 .The time course of R O2 describes medium re-oxygenation.Best-fit to a mono-exponential curve: R O2 = a À b 3 expðÀ P O2 3 tÞ estimates the O 2 permeability (P O2 ) of the oil barrier.This information is used to calculate O 2 ingress driven by oxygen-consuming cells.
Note: Curve-fitting can be performed using a package such as MATLAB or bespoke methods.Note: This step is omitted when using non-adherent cells.

KEY RESOURCES
d. Collect floating cells into 50 mL tube by washing the plate with fresh culture medium.e. Count the collected cells using hemocytometer or automated cell counter.f.Centrifuge cells at 400-600 g for 5 min at room temperature (20 C-24 C). g.Discard the supernatant and resuspend the pellet in standard culture medium, as deemed appropriate for the cell line of choice.
h. Aliquot the cell suspension across 1.5 mL sterile tubes.Set aside up to 5 additional tubes with lower cell concentrations for optimizing the gating protocol.
i. Place the tubes containing cells in an incubator at 37 C for up to 1 h prior to sorting.j.Prepare 1.5 mL cell-free collection tubes containing 500 mL of culture medium.k.Warm lactate-loading medium and lactate-free sorting medium and transfer sufficient volume into 50 mL tubes.
Note: A total of 18 tubes containing 3 3 10 6 cells each will require $25 mL per medium.
2. Prepare equipment for cell sorting: a. Follow the device manufacturer's recommended start-up procedure.v. Centrifuge at 400-800 g at room temperature (20 C-24 C). vi.Remove lactate-containing supernatant and resuspend the pellet in 1 mL lactate-free sorting medium.vii.Pass sample through cell-strainer.viii.Add 1 mg/mL DAPI and run the sample on the sorter using gates defined in Step 10(a).ix.Record events as soon as possible and at regular intervals thereafter.Cells producing the largest alkaline transients will have a higher 640/590 fluorescence ratio.
Note: Time points of 2, 5, 7, 10 and 12 min are recommended but may require optimizing in a cell line-specific manner.
x. Draw a gate around the most prominent alkaline population.
Note: This region corresponds to cells of the highest metabolic activity.

CRITICAL:
The proportion of cells falling into this region should not exceed 20%.d.Remove the supernatant and resuspend pellet in lactate-loading medium.e. Incubate for at least 5 min at room temperature (20 C-24 C).
CRITICAL: The next four steps (f-i) must be performed as rapidly as practical.
f. Centrifuge the tube at 400-800 g at room temperature (20 C-24 C). g.Remove lactate-containing supernatant and resuspend the pellet in lactate-free sorting medium.h.Strain through a cell-strainer.i. Add 1 mg/mL DAPI and run the sample through the sorter.
Note: Preparation of the next samples for sorting can begin at this point with cSNARF1 loading.j.Collect cells into 1.5 mL tubes.k.Replace with new tubes when cell count reaches 2 3 10 5 cells.
CRITICAL: Note the exact number of cells collected.
l. Repeat steps a-k for the remaining samples. 5. Optional: Seed the collected cells for metabolic phenotyping: a. Spin down the collected cells at 600 g for 5 min at room temperature (20 C-24 C). b.Gently aspirate the supernatant and resuspend cells in fresh culture medium, as appropriate for the cell type under investigation.c.Seed collected cells in equal numbers onto separate wells as soon as possible.
Note: Best results are achieved with seeding densities 30,000 to 80,000 cells/well.Note: Sterile, tissue culture-treated, black wall/clear bottom 96-well plates are recommended.If the cell yield is insufficient, 384-well plates with smaller wells can be used.
CRITICAL: Handle cells as soon as possible to avoid cell death.If a delay is unavoidable, place tubes with collected cells in the incubator.CRITICAL: When seeding cells, reserve the outermost wells of the plate for filling with PBS to reduce evaporation from innermost wells of the plate.CRITICAL: Include wells with medium but no cells to represent cell-free controls.

Metabolic phenotyping of sorted subpopulations
Timing: several hours, typically overnight, followed by off-line analysis time.Here: 1 h to prepare the plate for measurements, 17 h of fluorescence readings, 1 h of analysis.
Metabolic phenotyping uses a fluorimetric method to simultaneously measure medium pH and dissolved O 2 .This approach can be used to confirm sorting of cells by metabolic rate.Alternatively, the method can be used as a stand-alone assay.CRITICAL: Include cell-free wells to obtain a baseline for referencing metabolic activity.
b. Replace medium with 100 mL of fluorescent dye-containing low-buffering medium.c.Gently tilt the plate and dispense 150 mL of mineral oil over the wells by touching the upper walls whilst discharging the pipette.

Note:
The oil is a barrier to O 2 ingress, which is necessary for enabling respiration to meaningfully reduce dissolved O 2 .This barrier will also reduce CO 2 egress and contribute to medium acidification.
Note: Some wells can be left without an oil barrier to report acidification rate due to fermentation, without a component due to CO 2 hydration.

Note:
The duration of measurements can be as long as required: 18 h is recommended.
c. Export measurements and calculate R pH as F 2 /F 1 and R O2 as F 3 /F 4 .9. Run the analysis: a. Export the data into a workbook (e.g., Excel).See Data S1 for example.
CRITICAL: Implement one channel per sheet, with columns corresponding to wells and rows corresponding to time.
b. Prune columns that relate to empty or PBS-containing wells.c.Format the workbook using the example workbook as a template.
CRITICAL: To expedite analysis, ensure all sheet names and column headings in the ''sam-ple_ metadata'' sheet match (including case) those in example workbook.Columns should begin from A1 and there should be no gaps between rows or columns.In the sheets corresponding to channels (HPTS_400, HPTS_460, HPTS_isosbestic, and RuBPY), the first column should be time in the format ''hh:mm:ss''.In the sample metadata sheet, cell-free controls should be labeled as ''blank'' under the subpopulation column.
10. Run the analysis script in RStudio using code provided in Code S2. a. Open RStudio and create a new R Project directory.b.Within the directory, save a copy of analysis_script.Rmd and the formatted workbook.
Note: If necessary, install the packages listed in Section 1 of the analysis script according to system dependencies: tidyverse, readxl, and writexl.
c. Update the calculated HPTS and RuBPY calibration variables to Section 2 of the analysis script.
CRITICAL: The calibrations given in the scripts are exemplar and must be replaced with measured values applicable to the equipment used.
d. Update the constants specific to the experimental setup in Section 3 of the analysis script: i. volume of medium/well in mL (V), ii.calculated oxygen permeability constant (P O2 ).
e. Amend the file paths to the input Excel workbook in Section 4 of the analysis script.
f. Amend the file paths to the output Excel files and plots in Section 10 of the analysis script.

EXPECTED OUTCOMES
The sorting protocol separates cells by capacity to remove lactic acid, as determined by the pH signature of cells during an outward lactic acid gradient.According to our finding, 1 cells with higher lactate efflux capacity are associated with higher metabolic rate, both fermentative and respiratory, as confirmed using the metabolic phenotyping.The membrane impermeable ratiometric pH-dye HPTS tracks medium pH which provides a readout of metabolic rate, most of which is due to lactic acid production, a non-volatile acid.The rate of acid production is calculated from the product of pH change (measured during the experiment) and buffering capacity (measured separately, in cell-free experiments).Cumulative acid production is calculated from the sum of acid-production flux over time.RuBPY, normalized to the isosbestic (i.e., pH-insensitive) HPTS wavelength, provides a measure of dissolved oxygen.The rate of O 2 consumption by cells must consider the change in dissolved O 2 plus the rate of O 2 ingress from the atmosphere.The latter is calculated from the difference in O 2 between the medium and atmosphere multiplied by the O 2 permeability of the oil barrier.Following metabolic verification that the sub-populations are distinct, further experiments can be performed to understand the basis of metabolic heterogeneity.Exemplary data showing the time-course of transient populations collected by FACS and phenotyping data obtained with described protocol are shown in Figure 3.

QUANTIFICATION AND STATISTICAL ANALYSIS
The analysis of pH and O 2 time courses converts raw fluorescence reads into measurements of metabolic acid production and respiratory O 2 consumption, respectively.The computational pipeline for processing these data is provided in the R markdown file analysis_script.Rmd and may be run in RStudio.The main input required is a workbook which may be exported from a plate reader, and formatted using the example_workbook.xlsxas a template.In addition, calculated calibration variables and setup-specific constants must be updated within the script for accurate analyses.The main outputs are Excel files providing summary statistics of each time interval and sorted subpopulation, covering: medium pH, cumulative acid production (mmol), medium oxygen (%), cumulative oxygen consumption (mmol).
Examples of these outputs generated by running the analysis for example_workbook.xlsxare provided in the supplemental information.
Medium pH is derived from the ratio of deprotonated to protonated HPTS (R pH ) following Equation 1.
Changes in medium pH reported by HPTS represent the levels of free H + ions in the medium.However, chemical systems manifest pH buffering attributable to serum proteins and buffering agents such as HEPES and MES.Consequently, many H + ions generated by metabolism will be buffered.To describe the total metabolic H + flux, buffering must be accounted for using its quantitative measure, b.Thus, the flux of H + ions produced (J H ) is the product of the negative pH change (dpH/dt) and buffering capacity b.The sign is inverted because acid production reduces pH.Buffering is a function of medium pH, which we have measured previously for low-buffering medium containing 2 mM equimolar HEPES:MES (Figure 4).Cumulative acid production is calculated from the sum of H + fluxes (J H ) over time (from 0 to T) in cell-containing wells above cell-free controls.To convert concentration into molar amount, production is multiplied by medium volume (V), as shown in Equation 2. RuBPY fluorescence, as measured by F 4 , is quenched by oxygen.To provide a robust readout of medium oxygen, RuBPY is expressed relative to the pH-insensitive isosbestic point of HPTS fluorescence measured as F 3 .This exploits the fixed mixing ratio.The resulting ratio R O2 is converted into medium oxygen, in units of % of oxygen gas, as per Equation 3 using the RuBPY calibration variable r normoxia .
O 2 = 21 3 1 À 1 À R O2 =R normoxia 1 À r anoxia (Equation 3) In vivo, oxygen buffering can involve chelators, such as hemoglobin.However, oxygen buffering in media is close to zero and can be ignored.Oxygen is volatile and therefore two fluxes must be considered when measuring consumption: (i) depletion of medium O 2 and (ii) ingress of O 2 into medium from the atmosphere.The latter is reduced by the mineral oil barrier.To convert O 2 levels expressed as a percentage of gases (i.e., fractional partial pressure) into molar concentration, recordings are multiplied by solubility a, which can be obtained from recordings in water or saline 4 at the appropriate temperature, taken here as 10.6 mM per %.Cumulative O 2 consumption is calculated using Equation 4:

LIMITATIONS
The sorting protocol may not be suitable for cell lines that have insufficient metabolic heterogeneity or low population-averaged metabolic rate.The protocol yields a relatively low number of cells (<100,000) per sorting session which may not be sufficient for some types of measurements downstream.It is, however, possible to pool equivalent subpopulations from several rounds of sorting.
The experimental protocol was developed on a BD FACSAria III Cell Sorter and metabolism was Protocol sorted subpopulations cannot be resolved.To address this, cell seeding density should be increased or medium volume decreased.

RESOURCE AVAILABILITY
Lead contact Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact Pawel Swietach (pawel.swietach@dpag.ox.ac.uk).

Technical contact
Technical questions on executing this protocol should be directed to and will be answered by the lead contact.

Materials availability
This study did not generate unique reagents.

d.
Sterile filter using a 0.22 mm filter unit.Pause point: Store media at 4 C until use. 5. Prepare stock of fluorescent dyes: a. Dissolve RuBPY in sterile, deionized water (ddH 2 O) to make 100 mM stock.b.Dissolve HPTS in sterile, deionized water to make 4 mM stock.c.Mix RuBPY and HPTS stocks in a 1:1 v/v ratio.CRITICAL: Protect fluorescent dyes from direct exposure to light.Pause point: Fluorescent dye stocks can be stored for up to 2 months at À20 C. 6. Perform calibration of pH-and O 2 -sensitive dyes (Figure 1):

Figure 1 .
Figure 1.Schematic of protocol to calibrate HPTS and RuBPY fluorescence

7.
Calculate the permeability of the oil barrier for measuring O 2 consumption: Note: Respiratory rate is inferred from O 2 consumption, but an open system would rapidly equilibrate medium and atmospheric O 2 .To facilitate medium O 2 depletion, an oil barrier is placed on top of media to slow atmospheric O 2 ingress and allow metabolism to meaningfully change dissolved O 2 .The estimate of O 2 consumption must consider O 2 ingress, which is Protocol a product of permeability and gradient.The former depends on the volume of oil and can be calculated by imposing a O 2 gradient and measuring the rate of dissolved O 2 in the medium: a. Load plate with fluorescent dye-containing high-buffering medium at pH 7.4.Note: A 96-well plate and solutions at 100 mL per well are recommended, using at least triplicates per calibration point.b.Gently tilt the plate and add a volume of light mineral oil over wells to control the magnitude of the O 2 barrier.

4 .
Sort and collect cells separated by metabolic activity: a. Add cSNARF1 from stock to the cell suspension to a concentration of 10 mM. b.Mix and incubate at room temperature (20 C-24 C) for 10 min.c.Centrifuge tube at 400-800 g at room temperature (20 C-24 C).

Figure 2 .
Figure 2. Gating strategy Step 1: Draw a gate around the bulk of the cell population.Step 2: Exclude DAPI-stained (i.e., dead) cells to ensure that only living cells are analyzed and collected.Step 3: Exclude cell doublets.Step 4: Draw a gate that represents alkaline cells (i.e., cells of high lactate efflux capacity, corresponding to higher metabolic rate) and a gate around acidic cells (i.e., cells of low lactate efflux capacity, corresponding to lower metabolic rate).
6. Obtain media with dissolved pH-and O 2 -sensitive fluorescent dyes: a. Warm low-buffering medium to 37 C and thaw an aliquot of the HPTS/RuBPY stock.CRITICAL: Thoroughly vortex the HPTS/RuBPY stock mixture to ensure dyes are completely dissolved and evenly mixed.STAR Protocols 5, 103105, June 21, 2024 Protocol b.Dilute the HPTS/RuBPY stock mixture 1:1,000 (v/v) in low-buffering medium.7. Load 96-well (or similar) plate under sterile conditions: a. Once cells have become adherent, remove the plate from the incubator.

Figure 3 .
Figure 3. Exemplary data depicting metabolic heterogeneity of the MIA PaCa-2 cell line Panel A shows a series of scatter plots of pH in cells during evoked lactate efflux.The alkaline population (red; A) corresponds to the sub-population of cells with higher lactate efflux capacity, while the acidic sub-population (blue; B) describes cells with lower lactate efflux capacity.Note, the transient character of the alkaline sub-population.Panel B depicts the distinct metabolic phenotypes of collected subpopulations in terms of medium pH and medium oxygen, cumulative acid production and oxygen consumption.For clarity, a single representative repeat is shown.

Figure 4 .
Figure 4.The pH-sensitivity of buffering capacity of the low buffering medium CRITICAL: A suitable gas regulator is required to maintain gas composition, for example an Agilent dual CO 2 and O 2 gas controller.

TABLE REAGENT
(Continued on next page) STAR Protocols 5, 103105, June 21, 2024 (Continued on next page) STAR Protocols 5, 103105, June 21, 2024 Protocol STEP-BY-STEP METHOD DETAILS Sorting the metabolic subpopulations according to the lactate efflux capacity Timing: 2-3 h, depending on number of samples required This step separates and collects sub-populations of distinct metabolic activity, inferred from the magnitude of the intracellular pH (pHi) transient evoked upon lactate removal.1. Prepare cells on the day of sorting: a. Obtain culture vessels (e.g., plates) containing cultured cells.b.Aspirate media and wash with 1x phosphate-buffered saline (PBS).c.Add sufficient volume of 2x Trypsin-EDTA mixture to cover the plate and incubate at 37 C for 3 min until cells detach.
Cell-free wells, which maintain 21% O 2 , can be used to offset measurements, if necessary.