Cell freezing protocol optimized for ATAC-Seq on motor neurons derived from human induced pluripotent stem cells

In recent years, the assay for transposase-accessible chromatin using sequencing (ATAC-Seq) has become a fundamental tool of epigenomic research. However, it has proven difficult to perform this technique on frozen samples because freezing cells before extracting nuclei impairs nuclear integrity and alters chromatin structure. We describe a protocol for freezing cells that is compatible with ATAC-Seq, producing results that compare well with those generated from fresh cells. We found that while flash-frozen samples are not suitable for ATAC-Seq, the assay is successful with slow-cooled cryopreserved samples. Using this method, we were able to isolate high quality, intact nuclei, and we verified that epigenetic results from fresh and cryopreserved samples agree quantitatively. We developed our protocol on a disease-relevant cell type, namely motor neurons differentiated from induced pluripotent stem cells from a patient affected by spinal muscular atrophy.


Introduction
Since its establishment, the assay for transposase-accessible chromatin using sequencing (ATAC-Seq) has revolutionized the study of epigenetics 1 It is essential to preserve the native chromatin architecture and the original nucleosome distribution patterns for ATAC-Seq. Freezing samples prior to the purification of nuclei can be detrimental to nuclear integrity and chromatin structure, thus restricting the application of ATAC-Seq to freshlyisolated nuclei. This limits the use of ATAC-Seq on clinical samples, which are typically stored frozen, and represents a major logistical hurdle for long-distance collaborative projects, for which sample freezing is often inevitable.
In an attempt to overcome this drawback, we developed an optimized freezing protocol suitable for native chromatin-based assays. We tested two different freezing methods: flash-freezing and slowcooling cryopreservation. Flash-freezing is a procedure in which the temperature of the sample is rapidly lowered using liquid nitrogen, dry ice or dry ice/ethanol slurry, in order to limit the formation of damaging ice crystals. Conversely, slow-cooling cryopreservation lowers the temperature of the sample gradually and makes use of cryoprotectants, such as dimethyl sulfoxide (DMSO), to prevent ice crystal nucleation and limit cell dehydration during freezing. Cryopreservation techniques are widely employed for cell banking purposes and are routinely used in assisted reproduction technologies 4,5 .
We tested the freezing techniques using disease-relevant cell types, namely motor neurons (iMNs) differentiated from human iPSCs, which were derived from the fibroblasts of a patient affected by spinal muscular atrophy (SMA). This disease is caused by homozygous loss of the SMN1 gene and is characterized by the degeneration of lower motor neurons 6 .
We introduced a number of experimental quality control (QC) checkpoints and steps for data analysis to monitor the efficacy of the procedures and quantify potential alterations induced by cell freezing.
The method we describe should be applicable in a wide variety of settings and expand the number and types of samples that can be studied using ATAC-Seq.

Description of experimental design and overview of the protocol
We generated ATAC-Seq data on fresh (F), flash-frozen (FF), and cryopreserved (C) iMNs by following the procedure outlined in Figure 1. Fresh and frozen neurons were derived from the same pool of cells and processed in parallel in order to estimate the effects of freezing on ATAC-Seq outcomes without any batch effect bias.
The ATAC-Seq protocol was adapted from Buenrostro et al. 1,7 , with some modifications. Given that a successful ATAC-Seq experiment begins with the isolation of high-quality intact nuclei, we first introduced a quality control checkpoint consisting of the morphological evaluation of nuclei with either Trypan Blue or DAPI staining, followed by the accurate quantification of those nuclei using an automated cell counter. Precise counting of nuclei is important to ensure optimal tagmentation (the simultaneous fragmenting of the DNA and insertion of adapter sequences) and to limit the technical variability across samples. From a qualitative perspective, individual intact nuclei with a round or oval shape should be observed with no visible clumping. To exclude samples with severe degradation or over-tagmentation, we assessed the quality of the treated chromatin samples by gel electrophoresis, as introduced a size-selection step to enrich for nucleosome-free fragments. This step increases the signalto-noise ratio and improves the sensitivity of the methodology. After size-selection, libraries were PCR-amplified and submitted for single-end sequencing.

ATAC-Seq on iPSC-derived motor neurons (iMNs): flash-frozen cells
We first performed ATAC-Seq on fresh and flash-frozen iMNs. Differentiated neuronal cells were generated as described in Methods. We performed immunocytochemistry experiments using antibodies against markers of mature motor neurons to test the efficiency of the differentiation protocol; we showed that patient-derived iPSCs were successfully differentiated into ISL1-and SMI32-positive motor neurons ( Figure 2). Figure 3 shows ATAC-Seq outcomes from two representative samples.
Nuclei from fresh cells passed quality control, while nuclei from flash-frozen neurons exhibited excessive clumping, likely caused by disruption of the nuclear envelope and consequent leakage of DNA ( Figure 3A). After the transposase reaction, we assessed the quality of the resulting libraries by qualitative evaluation of agarose gel electrophoresis. The library from freshly-isolated nuclei displayed clear nucleosome phasing, while the library from flash-frozen neurons showed DNA smearing on the gel ( Figure 3B). This result strongly indicates that loss of chromatin integrity occurred during flashfreezing. We proceeded with next-generation sequencing for one fresh and one flash-frozen sample and uploaded the data on the UCSC Genome Browser for manual inspection of tracks and local visualization of peaks ( Figure 3C). As a negative control, we treated human naked DNA with the hyperactive Tn5 enzyme and sequenced this library alongside the ATAC-Seq samples. ATAC-Seq peaks from fresh neurons were sharp and overlap with H3K4me3 signals from ENCODE ChIP-Seq datasets. Using a MACS2 q-value threshold of 0.05, we obtained more than seventy thousand significant peaks using fresh cells. In contrast, the reads from flash-frozen cells were distributed evenly 6 across the entire genome, similar to the results obtained with the negative control, and less than five hundred significant peaks were detected. These findings indicate that flash-freezing of iMNs is not suitable for ATAC-Seq.

ATAC-Seq on iPSC-derived motor neurons (iMNs): cryopreserved cells
Next, we compared ATAC-Seq results from fresh and cryopreserved cells. Fresh iMNs were transferred to Cryostor media and slowly frozen, stored, and then thawed for processing (see Methods).
As shown in Figure 4A, nuclei from the cryopreserved cells were of high quality and the nucleosome laddering was detected by gel electrophoresis ( Figure 4B). Sequencing data from both fresh and cryopreserved samples showed sharp peaks and low background signal ( Figure 4C). Furthermore, the qPCR enrichment of the positive control site (GAPDH gene promoter, Figure 5A top panel) over the Tn5-insensitive site (gene desert region, Figure 5A bottom panel) was high and comparable to that of fresh cells, as opposed to qPCR results from flash-frozen neurons, for which less than 10-fold enrichment was observed ( Figure 5B). As in the case of fresh cells, we obtained more than seventy thousand significant peaks using cryopreserved samples (MACS2 q-value threshold = 0.05) ( Table 1).
These results reveal that slow-cooling cryopreservation is compatible with native chromatin-based epigenetic assays.

Quantitative comparison of fresh and cryopreserved iMNs
We subsequently performed a series of analyses to quantitatively compare the results from fresh and cryopreserved neurons. We generated sequencing data on three technical replicates from both conditions to assess whether the cryopreservation method induces any modifications in chromatin accessibility. All replicates originated from the same initial batch of cells. Information about 7 sequencing data for each sample is reported in Table 1. The percentage of reads mapping to the human genome was similar for all replicates, but cryopreserved samples displayed higher number of reads mapping to mitochondrial DNA. Despite this discrepancy, we proceeded with our analysis to assess the reproducibility of the epigenetic signal from nuclear DNA across all replicates. To this purpose, we removed mtDNA reads, normalized the libraries to have the same total read counts, and examined the number of reads in 5kb genome windows (excluding ENCODE blacklisted regions). Overall, we observed high reproducibility rates (R ≥ 0.978) between technical replicates in both fresh and cryopreserved samples ( Figure 6A). Remarkably, cryopreserved and fresh samples were almost as highly correlated to each other (R ≥ 0.973) as the technical replicates, which suggests that cryopreservation successfully preserves the read distribution across the genome. To further evaluate the similarity between cryopreserved and fresh samples, we identified the peaks in each sample and assigned each one of these peaks to neighboring features (promoters, exons, introns, distal intergenic regions and sites located downstream of the gene) within 1kb ( Figure 6B). The distribution of peaks with respect to features in the genome was highly similar across all samples, with most of the peaks located in intergenic regions and promoters. Next, to identify and quantify potential epigenetic alterations induced by the cryopreservation procedure, we performed analysis to detect sites that were significantly different between fresh and cryopreserved samples (see Methods). MACS2-derived peaks across all samples were merged into non-overlapping unique genomic intervals resulting in 75,711 sites. We then used edgeR to detect the differences between the two conditions. We identified very few differentially enriched sites across the genome (210 out of 75,711 total). The magnitude of the differences was small, never exceeding 3-fold (Figure 7). With the exception of chromosome 16, where no differentially enriched sites were identified, the sites spanned the genome, showing no obvious regional biases.
In conclusion, we established a cell freezing protocol suitable for ATAC-Seq experiments on iPSCderived motor neurons. As in the case of fresh samples, the cryopreserved samples passed all of the 8 quality control checkpoints. Although we observed that higher numbers of reads map to mitochondria DNA in cryopreserved iMNs, we demonstrated that the epigenetic signal from nuclear DNA was highly reproducible between fresh and cryopreserved neurons. This optimized protocol provides a simple and practical approach to extend ATAC-Seq to frozen cells. We anticipate that this work will be of great value to epigenetic investigators.

Motor Neuron Precursors (iMPs)
The SMA patient line, 77iSMA-n5, was grown until 90% confluent using a standard iPSC maintenance protocol. On Day 0 of differentiation, iPSCs were lifted as single cells by Accutase treatment for 5 min at 37ºC. We counted the cells and re-suspended them in Neuroectoderm differentiation media (NDM+LS), which contains 1:1 IMDM/F12, 1% NEAA, 2% B-27, 1% N2, 1% Antibiotic-Antimycotic, 0.2 µM LDN193189 and 10 µM SB431542. Next, we seeded 25,000 cells/well in a 384well plate and centrifuged the cells for 5 min at 200 rcf. On day 2, we transferred the neural aggregates to a poly 2-hydroxyethyl methacrylate (poly-Hema) coated flask and cultured them for an additional 3 days in NDM+LS media. On day 5, we seeded the neural aggregates onto a tissue culture plate coated with laminin (50µg/mL) to induce rosette formation.

Motor Neuron Cultures (iMNs)
We derived the iMNs by thawing the iMPs and immediately plating the single cell suspension onto plastic tissue culture-treated plates coated with 50µg/mL laminin for two hours at 37°C. We seeded the  Cryopreservation: pellets were re-suspended in Cryostor media and frozen slowly in a Mr. Frosty isopropyl alcohol chamber (FisherSci) at -80°C overnight. This procedure allowed us to achieve a rate of cooling of -1°C/minute. Both the flash-frozen isolated cell pellets and the cryopreserved iMNs were stored at -80°C until experimentation. To thaw the cryopreserved iMNs, we removed the cryovials from -80°C and quickly warmed them for 2 min in a 37°C water bath. We transferred the samples to 12 ml of warm 1X PBS supplemented with 1X protease inhibitor cocktail. We gently mixed each tube by inversion and centrifuged at 200 rcf for 5 min at 4°C. We carefully aspirated the supernatant and proceeded with nuclei isolation. Flash-frozen cell pellets were removed from -80°C and immediately re-suspended in 11 ice-cold cell lysis buffer.

Purification of nuclei from iMNs
We re-suspended the cell pellets in ice-cold cell lysis buffer (10 mM Tris-HCl, pH 7.4, 10 mM NaCl, 3 mM MgCl2, 0.1% IGEPAL CA-630) supplemented with 1X protease inhibitor cocktail (Roche). We incubated the cells on ice for 5 min and centrifuged at 230 rcf for 5 min at 4°C. We carefully removed the supernatant and re-suspended the nuclei in 25 µl of ice-cold 1X Tagment DNA Buffer (Illumina).
We quantified the nuclei with Trypan Blue staining and the Countess® Automated Cell Counter (Invitrogen).

DNA extraction
We purified the DNA from iMNs using the DNeasy Blood & Tissue Kit (Qiagen), according to the manufacturer's instructions. We quantified the DNA using a NanoDrop 2000 instrument (Thermo Scientific) and used 50 ng to prepare the DNA library using the Nextera DNA Library Preparation Kit (Illumina), according to the manufacturer's instructions. This library, obtained from naked DNA, was used as internal control to determine the background level of intrinsic accessibility of genomic DNA and correct for any Tn5 transposase sequence cleavage bias.

Chromatin tagmentation and sequencing
We used 50,000 nuclei for the transposase reaction, which was carried out as described in Buenrostro et al. 1

Bioinformatic analysis
We aligned sequencing reads to the hg19 genome build using BWA. We assessed the quality of the sequences using FastQC (more details on how the data was processed can be found at http://openwetware.org/wiki/BioMicroCenter:Software#BMC-BCC_Pipeline). Given the large percentage of mitochondrial reads found in some samples, we removed mitochondrial reads from the      The changes were small (< 3-fold) and no regional bias for these sites was observed. Although the percentage of mitochondrial DNA (mtDNA) contamination is higher in cryopreserved samples when compared to fresh samples, the number of significant peaks mapping to nuclear DNA is similar across all samples. (F = fresh, C = cryopreserved).