Protocol for single-cell ATAC sequencing using combinatorial indexing in mouse lung adenocarcinoma

Summary Single-cell ATAC sequencing using combinatorial indexing (sciATAC-seq) enables the identification of chromatin accessibility profiles at single-cell resolution with a dual-barcoding approach during transposition and library construction. Unlike commercial droplet-based approaches, sciATAC-seq is a cost-effective, extensible strategy that permits flexibility in the experimental scale via multiplexed barcoding across samples or perturbations. In contrast, droplet-based approaches have higher cell recovery and may be advantageous when cell input is limited. The step-by-step sciATAC-seq protocol described here is amenable to diverse cell types and fixed samples. For complete details on the use and execution of this protocol, please refer to LaFave et al. (2020).

into C3013 cells and incubate for 16 h at 37 C on Luria Broth (LB) plates with 100 mg/mL ampicillin. 8. Inoculate a 10 mL starter culture of LB media with 100 mg/mL ampicillin and incubate at 37 C in a bacterial shaker at 200 rpm for 16 h. 9. Add 10 mL of culture to a new flask of 1 L LB media with 100 mg/mL ampicillin pre-added. 10. Incubate in a shaker at 37 C at 200 rpm until the Optical Density (OD) reaches 0.6 using a spectrophotometer.
Note: This should take approximately 3-4 h but can vary based on the bacterial shaker used.
11. Chill culture on ice for 30 min once OD 0.6 is reached. 12. Add IPTG at a final concentration of 0.25 mM to induce Tn5 expression, then incubate the culture at 18 C while shaking at 200 rpm for 16 h.
Note: IPTG should be prepared fresh prior to use.
13. Collect pellet by centrifugation at 4000 3 g for 15 min at 4 C. 14. Remove the supernatant and flash freeze the pellet at À80 C for 30 min or longer. Resuspend the frozen pellet in 40 mL chilled HEGX buffer with 13 Roche Complete EDTA-free protease inhibitor tablet.
Optional: Add 10 mL of Benzonase nuclease to the resuspended pellet.
15. Lyse cells using a Bioruptor Plus sonicator with 50% duty cycles, keeping cells on ice until sufficiently lysed. Alternatively, use a French press or microfluidizer at appropriate settings to lyse the cells. 16. Centrifuge sonicated lysate at 30,000 3 g for 20 min at 4 C. 17. Add 1.05 mL of 10% PEI (pH 7) dropwise to the stirring lysate solution and centrifuge for 10 min at 9000 3 g at 4 C to remove precipitated E. coli DNA. 18. Prepare an Econo-Pac Chromatography Column by packing a 2 mL aliquot of a chitin slurry resin into a disposable column. 19. Prepare HEGX buffer. 20. Wash column with 30 mL HEGX buffer. 21. Slowly add the soluble fraction to the chitin resin, then incubate on a rotator at 4 C for at least 8 h. 22. Wash the column thoroughly with 40 mL HEGX buffer. 23. Elute chitin slurry with 10 mL of elution buffer (HEGX supplemented with 100 mM DTT) on a rotator at 4 C for 48 h. 24. Collect eluate and dialyze twice in 500 mL of Tn5 Dialysis Buffer. 25. Determine the concentration of Tn5 using an A280 measurement (an A280 of 1.0 equals 0.616 mg/mL or 11.56 mM Tn5). Concentrate dialyzed protein using an Amicon Ultra-4 Centrifugal Filter Units 30 K if required to an A280 of 2.5.

OPEN ACCESS
Note: As an additional quality control, purified Tn5 enzyme can be run on an SDS-page gel to confirm appropriate molecular weight and purity ( Figure 1).
26. Add sterile 100% glycerol 1:1 to create a final 50% glycerol stock of the purified protein.
27. Confirm activity of the transposase by performing a comparison to bulk ATAC sequencing experiment with the commercially available TDE1 transposome (Buenrostro et al., 2013). Load purified transposase with a pair of adapters as described later in this protocol. Perform tagmentation experiment on 50 ng purified genomic DNA, such as commercially available HeLa genomic DNA. 28. Quantify the number of cycles to reach 1/3 of the maximum fluorescent intensity (Buenrostro et al., 2015) and use these values to compare Tn5 efficiency (Troubleshooting 1). 29. Dilute prepared Tn5 to the appropriate concentration where transposition efficiency is comparable to that of the commercially available Nextera TDE1 transposome using Dilution Buffer. 100 mM ME-comp oligo /5Phos/C*T*G*T*C*T*C*T*T*A*T*A*C*A*/3ddC/ 50 mM 1 3 mL Note: FACS Buffer can be stored at 4 C for 2 weeks.

KEY RESOURCES
Note: Prepare fresh transposition buffer for each experiment.
Note: Nuclei Isolation Buffer can be stored at 4 C for 2 weeks.
Note: 23 Reverse crosslinking buffer can be stored at room temperature (25 C) for several months. Note: Any sciP1.X and sciP2.X primer can be used in this mix. The fragments will not be sequenced so the primers do not need to match the samples' barcodes.
Note: SYBR mix is purchased as a 10,0003 concentrate in DMSO. This stock should be further diluted to a 103 stock using water. Storage in a foil-covered 1.5 mL Eppendorf microcentrifuge tube at À20 C is advised. Make sure that qPCR master mix is well mixed before use.
Note: Scale up master mix depending on sample number.

Alternatives:
All standard reagents in the protocol can be replaced by comparable reagents from other manufacturers.
This protocol requires preparation and optimization of lab purified Tn5. Unloaded Tn5 is commercially available but may not be as cost effective for this protocol.
This protocol uses Lonza FlashGels TM to image the ATAC fragment sizes. Samples can also be run on standard agarose gels. This protocol uses a BD FACSAria III sorter, but any comparable cell sorter can be used.

Reagent Final concentration Amount
This protocol uses a ViiA7 Real-Time PCR system, but any comparable qPCR machine can be used.

STEP-BY-STEP METHOD DETAILS
This protocol includes a number of adaptions and optimization from prior methodologies , Cusanovich et al., 2015, Lake et al., 2018, Pliner et al., 2018, Preissl et al., 2018.

Sample preparation
Preparation of samples will depend on sample collection methodology (Figure 2, Day 1 overview). For cell lines, we recommend detachment of cells with trypsin or accutase and then proceed to fixation and transposition. Fresh cells should be isolated or viably frozen cells should be thawed immediately and washed 1-2 times with 13 PBS before the start of the protocol. This protocol describes the specific steps for using fresh murine lung tumor samples . Lung tumors are isolated from Kras G12D/+ ; Trp53 -/-; Rosa26 tom/+ mice and dissociated cells are sorted for tdTomato expression to isolate tumor cells for sciATAC-seq. Tumors are induced at 6-8 weeks of age and late stage tumors are collected at 25 weeks post-tumor induction . Cell preparations can be used fresh or viably frozen in media (DMEM + 10% FCS) and 10% DMSO; however, we recommend performing the sort of the day of the experiment on viably frozen tumor cells.

Collection of lung adenocarcinoma cells
Timing: [4 hrs] a. Euthanize mouse using an isoflurane chamber. b. Dissect whole tumor-burdened lung or individual tumors and place in 1.5 mL Eppendorf microcentrifuge tubes containing 100 mL Lung Digestion Buffer. Buffer is based on manufacturer's instructions (https://www.miltenyibiotec.com/US-en/products/lung-dissociation-kitmouse.html#gref). c. Thinly mince tissues using scissors and then add an additional 0.5 mL Lung Digestion Buffer to tube. d. Incubate at 37 C with rotation for 30 min. e. Grind sample over a 100 mM filter into a 50 mL conical tube with pipette tip. f. Pipette 5 mL S-MEM over the filter to wash remaining sample into the tube. g. Pellet cells by centrifugation at 300 3 g for 5 min.
Note: Centrifugation speed and time may be adjusted according to sample type. h. Remove supernatant and resuspend pellet in 1 mL ACK lysis buffer to lyse red blood cells.
Note: Use of a dead cell removal kit (e.g., Miltenyi cat no. 103-090-101) is recommended. k. Sort for APC neg (immune depleted), tdTomato pos , and DAPI neg into FACS buffer using a FACS Aria (example gating scheme, Figure 3). Proceed to fixation and transposition.

Assemble Tn5
Timing: [45 min] a. Prepare Dilution Buffer, then cool on ice for 5 min. The dilution buffer can be stored at À20 C for several months. CRITICAL: Assemble Tn5 immediately before fixation and transposition steps. b. Prepare 13 Tn5 by diluting homemade Tn5 in Dilution Buffer to a final volume of 96 mL. c. Mix diluted Tn5 1:1 with pre-annealed adapters/glycerol mix by combining 4 mL diluted Tn5 with 4 mL each pre annealed adapter (20 adapters in total).
Note: For 96 transposition reactions, 8 sciAD1.X and 12 sciAD2.X distinct barcodes are used, see Table 4 for barcode combinations.
d. Incubate on the bench at 25 C for 30 min. e. After 30 min, proceed to fixation and transposition steps but keep on ice.
CRITICAL: Always coat tubes with 7.5% BSA before spinning to improve cell recovery. Add BSA to coat the entire tube and remove prior to cell fixation.

CRITICAL: Always use a swing bucket rotor for centrifugation instead of a tabletop centrifuge.
Note: Centrifugation time and speed may need to be adjusted for sample type based on prior experience. b. Resuspend pellets in 100 mL cold 13 PBS. c. Count cells by mixing 5 mL of each sample to 5 mL of trypan blue using a hemocytometer.
Note: Visually check samples for clumping and debris which may cause issues in future steps (Troubleshooting 2). d. Dilute each sample to 100K cells in a final volume of 100 mL.
Note: If cell concentration cannot be achieved, resuspend the entire cell sample in 100 mL. Variations to cell number must be accounted for at the PCR barcoding step by maintaining an appropriate cell density per reaction. e. Pellet cells by centrifugation at 300 3 g for 3 min. f. Dilute 16% formaldehyde to 1.6% with H 2 O. g. Fix cells by adding 6.7 mL 1.6% formaldehyde (final concentration 0.1%). Mix samples by pipetting (Troubleshooting 3).
Note: Fixation conditions have been optimized in specific human and mouse cells lines and have been successful in several tissues including lung and brain. Fixation percentage can be tuned from 0.1%-0.5% depending on cell fragility.
Note: Omitting the fixation step may enable experiments with limited cell number, but will increase the level of crosstalk between single cells due to excess debris in the sample prep (LaFave et al., 2020). h. Incubate on the bench at 25 C for 5 min. i. Prepare a master mix of 5.6 mL 2.5M glycine, 5.0 mL 1 M pH 8.0 Tris, and 1.3 mL 7.5% BSA for each 100 mL fixed sample. Stop the fixation by adding 11.9 uL master mix to each individual sample. j. Incubate on ice for 10 min. k. Gently wash the cells with 0.5 mL of 13 PBS by pipetting against the side of the tube without resuspending the pellet. Spin at 500 3 g for 3 min.

Transposition
Timing: [1.5 h] a. Prepare Transposition Buffer and Nuclei Isolation Buffer. b. In a 96 well plate, combine 7 mL of the transposition buffer and 1 mL fixed cells in each well. c. Incubate on the bench at 25 C for 10 min. d. Dilute the assembled Tn5 1:1 by adding 8 mL Transposition Buffer to 8 ml assembled Tn5. Add 1 mL diluted Tn5 containing sciAD1.X oligo and 1 mL diluted Tn5 containing sciAD2.X oligo to each well based on plate map (Table 4). e. Shake at 300 rpm for 30 min at 37 C using the Eppendorf ThermoMixer C. f. Quench the reaction by adding 1 mL of 0.5M EDTA to each well. Mix well by pipetting. g. Shake at 300 rpm for 15 min at 37 C.
Combine transposition reactions and prepare for PCR barcoding Transposed samples are pooled together and re-aliquoted into individual wells to add another level of barcoding at PCR amplification.  Note: 1 mL of RCB is required for each PCR plate. To increase recovery of cells, the protocol can be scaled up to 9 individual 96 well PCR plates. c. Create master mixes for each sciP1.x primer and each sciP2.x primer to aliquot in a 96 well plate.

Pool transposition reactions
Note: An example of how to aliquot primers across the plate can be found in Table 5. d. Prepare Master Mix 1 with the transposed cell sample (concentration 13.3 cells/mL determined above) and 10 mM sciP1.X for each row. To do this, add 15 mL of the transposed cells to 5 mL of each primer sciP1.X (12 total). e. Prepare Master Mix 2 with 10 mM sciP2.X and RCB with proteinase K for each column. To do this, add 50 mL RCB + proteinase K to 10 mL each primer sciP2.X (8 total).
Note: Prepare Master Mix 1 and Master Mix 2 in PCR tubes to allow for multichannel pipetting when distributing across the 96 well plate. f. Distribute 2 mL of Master Mix 1 across the 96-well plate according to the example in Table 5. g. Distribute 3 mL of Master Mix 2 across the 96-well plate according to the example in Table 5.
Note: Each well should have 2.5 mL RCB with proteinase K, 0.5 mL 10 mM sciP1.X, 0.5 mL 10 mM sciP2.X, and 1.5 mL transposed cells. Master Mix 2 is made in excess and leftover is expected.
Note: In the nomenclature for PCR barcodes, sciP1.X and sciP2.X, X refers to the full repertoire of barcoding space. h. Incubate in a thermal cycler at 55 C for 1 h to 16 h.
Pause point: Shorter incubation times from 1-16 h can be used. We recommend 16 h to provide a convenient stopping point between Day 1 and Day 2 of the protocol.

PCR amplification and quantification
Addition of barcodes through PCR and quality check prior to sequencing (Figure 4, Day 2 overview).

PCR Amplification
Timing: [45 min] a. Remove PCR plate from thermal cycler and briefly spin down plate at 300 3 g for 10 seconds. b. Quench reaction by adding 5 mL 10% Tween20 to each well. Mix well by pipetting using a multichannel pipette. c. Prepare PCR master mix by combining 1.25 mL 23 NEBNext PCR mix and 250 mL nuclease free H 2 O. d. Add 15 mL of the PCR mix solution to each well using a multichannel pipette. Pipette to mix and spin down plate. e. Perform PCR reaction as follows:

Quantitative PCR
Timing: [2 h] a. After the initial 5 cycles of amplification, remove the plate from the thermal cycler and keep on ice for the duration of this step. b. For a quality check, randomly choose 4-8 wells to use for quantitative PCR (qPCR) reactions to determine the number of additional cycles needed for the entire plate. A negative control using water instead of DNA is advised. c. Prepare qPCR master mix. Scale up based on the number of wells chosen to test above. d. Add 9 mL qPCR master each well of qPCR plate. Add 1 mL sample to each well. e. Perform qPCR reaction to saturation as follows: g. Perform additional cycles as follows: h. Run qPCR reactions on a Lonza FlashGel TM DNA cassette 2.2% gel for 9 min at 250 Volts. i. After running, the gel is imaged in a Biorad Gel Doc XR or similar UV trans-illuminator.
Note: Confirm that you see the expected fragmentation pattern following ATAC-seq (Figure 6). Failed experiment will appear as primer dimers and means that the experiment was not successful (Troubleshooting 4). 10. Next-generation sequencing and data processing a. Libraries can be sequenced on the Next-seq platform (Illumina) using a 150-cycle kit without custom sequencing primer based on the below barcoding strategy (Figure 8). The following read lengths must be used: Read 1: 47 cycles, Index 1: 36 cycles, Index 2: 36 cycles, Read 2: 47 cycles.

Quantification of minimally amplified libraries
Note: We suggest to sequence 20-40K reads per cell, but the number of required reads may vary depending on the sample quality. b. Convert base calls to fastq format using bcl2fastq. c. Trim sequencing reads to remove adapter sequences using trim_galore <filename(s)>. trim_galore automatically detects Nextera adapter sequences and trims low quality reads. d. Align reads to hg19 or mm10 genome using Bowtie2 (Langmead and Salzberg, 2012) using maximum fragment length set to 2 kb and all other default settings (bowtie2 -X2000 -rg-id). e. Demultiplex tolerating one mismatch base within barcodes. f. Remove mitochondrial and low-quality reads using SAMtools (Li et al., 2009) (samtools view -b -q 30 -f 0x2). g. Remove duplicate sequences with picard toolkit (http://broadinstitute.github.io/picard/) h. Call peaks from a single alignment (.bam) file using input peak calling with MACS v2.1.2 (MACS2). Use default options with the following flags set: -nomodel, -nolambda, -keepdup all, -call-summits returning a list of single base pair peak summits with FDR q < 0.01.
Note: If calling peaks across multiple experiments, aggregate each sample into one .bam file. Call peaks on each sample to obtain sample-specific peaks. i. Identify a list of significant, non-overlapping fixed width peak windows by padding peak summits with 150 base pairs at either side to generate 301 base pair window peak regions. j. Sort peaks in decreasing order of significance scores. Remove peak windows with the lower significance scores and keep the most significant peak. Repeat over an iterative process, to identify 301 base pair disjoin peak windows. k. Using the generated peak list above, determine the number of reads overlapping a given peak window for each unique cell barcode. l. Generate a peak X cell counts matrix which corresponds to ATAC reads in peaks for each cell profiled.

EXPECTED OUTCOMES
Following data processing, chromatin accessibility data can be then directly used in ATAC pipelines such as ChromVAR (Schep et al., 2017). For quality control, we propose a first pass pseudobulk check of transcriptional start site (TSS) enrichment using the ENCODE definition of annotated TSSs compared to background (below 8 is a failed library, 8-10 is acceptable, above 10 is good) (https://www.encodeproject.org/atac-seq/). Appropriate nucleosome patterning should also be detectable from aggregated pseudobulk data (Troubleshooting 6). We recommend assessing The barcoding strategy has been updated to avoid sequencing with custom primers. sciAD1.X, sciAD2.X, sciP1.X, and sciP2.X ultimately assign four unique barcodes (bc1-4) per cell. Read 1 (R1) and Read 2 (R2) are flanked by the Tn5 mosaic end (ME) transposition recognition sequences.

OPEN ACCESS
quality control measures for single-cell data prior to downstream analyses including using cut-offs of fragment of reads in peaks (FRIP) R 0.4 and a minimum of 1000 unique nuclear reads per cell for downstream analyses (Troubleshooting 7). Description of additional analysis tools can be found in LaFave et al. 2020.

LIMITATIONS
We have optimized this combinatorial indexing protocol to work well for multiplexing primary tumor samples, with improvements to avoid FACS sorting and improve fixation and transposition conditions. We recommend using 100K cells for combinatorial indexing with fixation. Low cell numbers may be challenging as there may be cell-type specific loss from fixation and serial washes. Adjustments for cell number should be made at all steps if using a reduced cell number. For example, the fixation step may be omitted to reduce cell loss and fewer transposition reactions may be needed to maintain a ratio of 1,000 cells per reaction. High sample viability at the start of the protocol is important to generate high quality data and cellular dissociation strategies should be optimized for distinct tissue types. There are no stopping points on the first day of the experiment before reverse crosslinking, which can be challenging when coupled with samples requiring long isolation times. Therefore, if sample preparation is long, we suggest viably freezing cells prior to beginning the first day of the protocol.

TROUBLESHOOTING Problem 1
In house Tn5 does not have comparable activity to commercial Tn5 (Before you begin).

Potential solution
An error in the protocol has likely altered the activity of Tn5. In order to identify which step is causing activity loss, it is recommended to assess Tn5 activity by bulk ATAC-sequencing at various steps in the protocol post cell-lysis. For example, perform bulk ATAC-sequencing before and after E. Coli precipitation to determine if there was activity loss.

Problem 2
Cells or transposed nuclei appear to have excess debris (step 3).

Potential solution
Add in a live-dead sort or dead cell removal approach to improve cell quality. Excess debris can impact cell quality as assessed by quality control metrics like FRIP.

Problem 3
Cells do not pellet well post-fixation (step 3).

Potential solution
Confirm that all steps used tubes coated with 7.5% BSA and a bucket rotor centrifuge was used. Primary samples of diverse cell-types can be challenging to pellet without pre-coating tubes with BSA.

Problem 4
Libraries appear on gel at a lower molecular weight (step 8i).

Potential solution
This may be indicative of over-transposition of too few cells. Data processing might result in usable libraries; however, increasing the number of cells as suggested in this protocol should overcome this issue.

OPEN ACCESS
STAR Protocols 2, 100583, June 18, 2021 Problem 5 Following qPCR, some wells did not amplify product above negative control. qPCR products only show primer dimers when run on a gel (Step 8).

Potential solution
Errors at several steps could result in loss of material. For example, cell counting may have been inaccurate and fewer cells than expected were added to each well at the beginning of the protocol or at cell splitting. Unequal distribution of cells across well could also be problematic. Confirm accurate cell counting and single-cell resuspension by counting cells at appropriate steps in the protocol.

Problem 6
Analyzed data appear to have low quality-control scores (step 10).

Potential solution
Fixation conditions may not have been optimized for sample type. Perform test experiments to optimize fixation conditions from 0.1%-1% to confirm data quality and representation of expected cell types in experiment.

Problem 7
Few cells pass filter after sample demultiplexing (step 10).

Potential solution
First, confirm that all experimental quality control checks (agarose gel fragmentation pattern and qPCR) look correct. Next, confirm that appropriate barcodes were used in the demultiplexing part of the protocol. Incorrect barcode assignment can remove high-quality data from the protocol. If barcoding demultiplexing is correct, confirm that peak file is correct (can spot check by assessing peaks in the aggregated .bam file). If an error is not detectable from computational analyses, revisiting debris elimination protocols may be helpful.

RESOURCE AVAILABILITY
Lead contact Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Lindsay LaFave (lmlafave@mit.edu).

Materials availability
New materials were not generated for this study.

Data and code availability
The data and code can be found online at https://www.sciencedirect.com/science/article/abs/pii/ S153561082030310X.

ACKNOWLEDGMENTS
This work was supported in part by grant P01-CA42063 from the National Institutes of Health, the Cancer Center Support (core) grant P30-CA14051 from the National Cancer Institute, and by the Howard Hughes Medical Institute (HHMI). We thank the flow cytometry cores at the Swanson Biotechnology Center, the Walk-up Sequencing Core at the Broad Institute, and the Bauer Sequencing

DECLARATION OF INTERESTS
T.J. is a member of the Board of Directors of Amgen and Thermo Fisher Scientific and a co-Founder of Dragonfly Therapeutics and T2 Biosystems. T.J. serves on the Scientific Advisory Board of Dragonfly Therapeutics, SQZ Biotech, and Skyhawk Therapeutics. T.J. is also the President of Break Through Cancer. His laboratory currently receives funding from Johnson & Johnson and Lustgarten Foundation, but these funds did not support the research described in this manuscript. J.D.B. holds patents related to ATAC-seq and scATAC-seq and serves on the Scientific Advisory Board of CAMP4 Therapeutics, seqWell, and Celsee.