HiChIP and Hi-C Protocol Optimized for Primary Murine T Cells

The functional implications of the three-dimensional genome organization are becoming increasingly recognized. The Hi-C and HiChIP research approaches belong among the most popular choices for probing long-range chromatin interactions. A few methodical protocols have been published so far, yet their reproducibility and efficiency may vary. Most importantly, the high frequency of the dangling ends may dramatically affect the number of usable reads mapped to valid interaction pairs. Additionally, more obstacles arise from the chromatin compactness of certain investigated cell types, such as primary T cells, which due to their small and compact nuclei, impede limitations for their use in various genomic approaches. Here we systematically optimized all the major steps of the HiChIP protocol in T cells. As a result, we reduced the number of dangling ends to nearly zero and increased the proportion of long-range interaction pairs. Moreover, using three different mouse genotypes and multiple biological replicates, we demonstrated the high reproducibility of the optimized protocol. Although our primary goal was to optimize HiChIP, we also successfully applied the optimized steps to Hi-C, given their significant protocol overlap. Overall, we describe the rationale behind every optimization step, followed by a detailed protocol for both HiChIP and Hi-C experiments.


Introduction
Understanding chromatin architecture is becoming the next gold standard when studying processes inside eukaryotic cells. Since the advent of the chromosome conformation capture technique [1], there has been a vast development of methods allowing to probe long-range chromatin interactions on a genome-wide scale [2][3][4]. Such widely used research approaches are Hi-C [5], or its variants and methods that enrich for proteinmediated long-range chromatin interactions, such as ChiA-PET [6,7], PLAC-seq [8] and HiChIP [9]. In this work, we describe the key steps of an optimized HiChIP protocol that can also be used to prepare Hi-C libraries since both techniques share a significant part of the protocol steps. The whole approach is based on the in situ version of the Hi-C protocol [10], where all the initial steps take place in intact nuclei to reduce the probability of generating chimeric DNA molecules as a result of ligating DNA fragments originating from different cells. The protocol's outline is depicted in Figure 1a. Briefly, a suspension of single cells is crosslinked, and the nuclei are isolated. Nuclear DNA is enzymatically digested, and the resulting 5 overhanging ends are filled in with biotinylated nucleotides to blunt and label them. The resulting DNA ends are ligated together, aiming to generate chimeric ligation products between two genomic fragments that were in close spatial proximity within the cell nucleus. Then, the nuclei are lysed, and the ligated DNA is fragmented by sonication. In the case of a Hi-C experiment, these fragments are directly used to prepare a sequencing library. In a HiChIP experiment, a protein of interest is further pulled down, as HiChIP protocol performed in our laboratory yielded a high number of fragments with dangling ends. The improved version of the protocol provided in this work substantially decreased the number of invalid pairs, including the dangling end pairs. Comparison with the publicly available CTCF HiChIP experiment [15] is provided.

Experimental Design
Both Hi-C and HiChIP experiments are performed in the same way up until the immunoprecipitation step. In this protocol, we used primary murine thymocytes, but we encourage readers to test it out also in other cell types. In HiChIP experiments, we recommend using 10 million cells to ensure sufficient yield of DNA after the immunoprecipitation step. A higher number of cells may decrease the efficiency of enzymatic reactions. For this reason, the fill-in and ligation steps were optimized for 5 million cells; hence each sample of 10 million cells was split in half. In Hi-C experiments, 5 million cells are enough as a starting point for high-quality DNA library preparation, so it is not necessary to split the sample in any step.

Cell Lysis and Digestion
The prerequisite for a successful experiment is to sufficiently lyse the cell membrane followed by isolation of the crosslinked nuclei. We always monitor the success of nuclei isolation, as previously suggested [16], by checking the integrity of cell membranes upon cell lysis, with trypan blue staining and microscopic evaluation. The insufficient isolation of nuclei would decrease the efficiency of the next steps, while extended lysis could dramatically decrease the yield of intact nuclei to be analyzed. All operations with isolated nuclei must be carried out with care not to damage them. Once the nuclei are isolated, they are treated with SDS, which further increases the risk of breaking them. This treatment ensures that the nucleus and chromatin are more approachable for the restriction enzyme. The selection of the proper restriction enzyme depends on the specific research question; however, in Figure 2a, we present the DNA fragment size distribution upon in silico digestion of the murine genome, by either MboI or HindIII, as examples of four and six base-pair cutters, respectively. Although MboI was successfully used in the pilot HiChIP study [9], it is supposedly sensitive to CpG methylation, so it is recommended to use its isoschizomer DpnII [17]. Digesting genomic murine DNA side by side with these two enzymes, we could not detect any difference in the relative sizes of DNA smear (Figure 2b), but there may be some differences at a finer resolution. To optimize the step of restriction enzyme digestion, we compared different durations, amounts of restriction enzyme and the concentration of Triton X-100. Triton X-100 at this step is primarily used to quench the SDS; however, it was also shown to positively impact the efficiency of restriction enzyme digestion [18]. Indeed, the increase in Triton X-100 concentration from 1% to 2% provided better digestion results (Figures 2c and A1a). This step is quite important since insufficient restriction enzyme digestion will decrease the efficiency of further steps.

Fill-in, Biotinylation and Ligation
In the next part of the protocol, the resulting 5 -end overhangs are filled in with nucleotides where one kind of nucleotides is biotinylated to label the end-to-end junction for the subsequent enrichment step. The length of the linker that attaches the biotin moiety to a specific nucleotide is reflected in the efficiency of its incorporation by DNA polymerase. The longer the linker, the less efficient the incorporation of the modified nucleotide is, but at the same time, the more efficient the recognition and biotin pull-down in the subsequent steps are. A 16-atom linker represents a compromise between these two aspects. In our experiments, we used biotinylated dNTPs, although biotinylated dATPs can also be used [17]. To optimize the biotinylation and ligation steps of the protocol, we performed a set of experiments with one condition being different at a time. For the optimization of the fill-in step, all the subsequent DNA ligation reactions were performed under the same conditions to track the fill-in efficiency, based on the efficiency of ligation for each experimental setup ( Figure A1b). Similarly, a set of identical fill-in experiments was performed, followed by a ligation step, performed under different experimental conditions, to identify the most efficient ligation setup ( Figure A1c). was enriched by SDS and Triton X-100 as indicated to mimic the HiChIP protocol conditions. (d) Comparison of selected ligation conditions revealed that condition L4 was the best performing. Three controls, Cdig, Cfill and Clig, represent samples for which enzyme in the indicated step was not added. Moreover, four controls, Nnuc, Ndig, Nfill and Nlig, were used for which no enzymes were used at any step, and samples were collected at the indicated step. These controls showed that the temperature changes and extended incubations did not affect chromatin integrity. (e) The optimized protocol was highly reproducible, as demonstrated by the comparison of control samples from two biological replicates from three different genotypes. Three controls provided represent the size distribution of isolated DNA after the digestion step, ligation step and after sonication, thus prior to the chromatin immunoprecipitation step. The DpnII digestion buffer (NEB, B0543S) was enriched by SDS and Triton X-100 as indicated to mimic the HiChIP protocol conditions. (d) Comparison of selected ligation conditions revealed that condition L4 was the best performing. Three controls, Cdig, Cfill and Clig, represent samples for which enzyme in the indicated step was not added. Moreover, four controls, Nnuc, Ndig, Nfill and Nlig, were used for which no enzymes were used at any step, and samples were collected at the indicated step. These controls showed that the temperature changes and extended incubations did not affect chromatin integrity. (e) The optimized protocol was highly reproducible, as demonstrated by the comparison of control samples from two biological replicates from three different genotypes. Three controls provided represent the size distribution of isolated DNA after the digestion step, ligation step and after sonication, thus prior to the chromatin immunoprecipitation step.
To prevent any inhibitory effects of the digestion buffer and/or residual SDS on the fill-in efficiency, we spun down the nuclei and discarded all the supernatant. In the fill-in and biotinylation steps, we mainly focused on the identification of the optimal temperature, time and amount of the enzyme. The large fragment of DNA polymerase I (Klenow) is used in this reaction to generate blunt DNA ends. The goal is to maximize the reaction efficiency to overcome the hurdles of biotinylated nucleotides incorporation. However, the Klenow fragment still retains its 3 to 5 exonuclease activity, which competes with the polymerase activity [19] and that can, under certain conditions, even digest blunt DNA ends [20]. Given the increased amount of enzyme needed in the HiChIP experiment together with the problematic biotinylated substrate, the exonuclease activity of the enzyme needs to be considered. The mutated Klenow fragment lacking its 3 to 5 exonuclease activity creates a single nucleotide overhang [21], which would inhibit the ligation reaction; thus, it cannot be used either. This fact is probably reflected in our data as the best result was achieved after only 30 min of treatment while both increased time and amount of enzyme displayed worse performance ( Figure A1b). Similar results could have probably been achieved in lower temperatures and extended incubation times, but we did not test them. To overcome the 3 to 5 exonuclease activity, we also incorporated non-heat inactivation of the Klenow enzyme by 0.5% SDS treatment. To overcome the inhibitory effect of SDS, we quenched it with Triton X-100, which also made the nuclei pellet visible upon centrifugation.
Before the ligation step, the nuclei should be again spun down and the supernatant discarded, as the presence of the residual dATP could compete with the ATP necessary for the ligation reaction [22]. For the ligation part, we again optimized the time, temperature and amount of the enzyme, but this time we also compared the effect of several additives to make sure that the blunt ends were efficiently joined. First, we tested polyethylene glycol (PEG), as a known crowding agent facilitating many enzymatic reactions and that is also used in the quick ligation kits. It has been previously shown to dramatically increase the ligation efficiency [23][24][25], depending on its concentration and molecular weight. As another crowding agent, we tested dimethylformamide (DMF), and last, we investigated spermidine, which was also shown to improve the ligation efficiency in the past [26]. The most efficient combination appeared to be either 1% PEG 6000 or 10% DMF together with a 6 h incubation at room temperature (Figures 2d and A1c).
The established protocol is highly reproducible, as demonstrated by the three-control comparison (before ligation, after ligation and after sonication) from samples of three different genotypes and from two biological replicates (Figure 2e).

Optimization of DNA Library Preparation
The first part of the protocol is applicable for both Hi-C and HiChIP experiments, and although it is optimized for T cells, it could also be applicable for other cell types. The next steps of the HiChIP experiment, e.g., nuclear lysis, shearing and immunoprecipitation, are highly dependent on the cell type used, and they should be optimized accordingly. In this work, we provide a protocol optimized for murine T cells.
The DNA library preparation steps are again universal since they employ the manipulation of pure DNA. In this protocol, we describe the DNA library preparation by the Nextera Kit from Illumina, which is easy and convenient. To lower the costs, it can alternatively be replaced by the use of a custom-made Tn5 enzyme [27]. However, traditional methods based on adaptor ligation are also possible. In that case, we recommend performing an additional size selection step before the library construction, followed by end-repair and biotin pull-down as previously described [17].
The Tn5 enzyme tends to stay tightly bound onto DNA [27,28], which would inhibit the downstream PCR amplification of DNA or at least the first "gap-filling" step required upon tagmentation. For this reason, we sought for a solution to strip away Tn5 with the unwanted fragments of DNA while keeping the biotinylated fragments bound on the streptavidin beads. We found that up to 0.3% final concentration of SDS, which is sufficient to strip away Tn5 [28], would still preserve binding of the biotinylated DNA fragments onto the streptavidin beads, providing a similar efficiency to the original HiChIP protocol, which lacks the SDS treatment (Figure 3a). Tn5 and non-biotinylated DNA fragments are then washed away, and the beads with purified ligation junctions are used for PCR amplification. The efficiency of biotinylation and theoretically even the stripping step can also be monitored by comparing the amount of input DNA and the amount of DNA in the supernatant (Figure 3b). Upon PCR amplification, we usually perform clean-up and size selection of DNA using AMPure beads. In this protocol, we provide AMPure beads to sample ratios that usually work well for us. However, this step can be optimized for each batch of beads by purifying 100 bp agarose gel ladder with different AMPure beads to ladder ratios to identify the optimal one. An example of the fragment size distribution of the working HiChIP library is provided in Figure 3c.    Spin down the cells at 500× g at 4 • C for 5 min and wash them twice with 10 mL of 1× PBS.

3.
Count the cells (for example, using a hemocytometer or an automated cell counter) and use 10 million cells per HiChIP experiment or 5 million cells per Hi-C experiment.
Bring up the volume to 10 mL with 1× PBS and resuspend the cells well.

5.
Incubate the samples for 10 min at room temperature while rocking. 6.
Quench the reaction by adding Glycine to 0.2 M final concentration (870 µL, 2.5 M stock), mix well and incubate at room temperature for 5 min. 7.
Spin down the cells at 1000× g at 4 • C for 5 min. 8.
Discard the supernatant, resuspend the cell pellet in 1 mL ice-cold 1× PBS and transfer the solution into a fresh Eppendorf tube. 9.
Spin down the cells at 1000× g at 4 • C for 5 min. 10. Repeat the wash (steps 8-9) one more time and discard the supernatant.
32. Merge the two halves of the HiChIP samples back together (does not apply to Hi-C), mix carefully by pipetting and take 120 μL as the control C2.
33. Spin down the nuclei at 2400× g at 4 °C for 15 min and discard the supernatant.
34. To further decrease the number of dangling ends, we recommend implementing the removal of unligated, biotinylated ends [17]. For HiChIP libraries, it can either be performed right after the ligation in T4 ligation buffer (step 32) or after nuclei spin down (at this step, in its dedicated T4 ligase reaction buffer). For the reaction, add only dGTP and dATP nucleotides and T4 DNA polymerase, which will naturally manifest its 3′-5′ exonuclease activity and remove the unligated 3′ biotin-dCTP. Alternatively, this step can be performed after DNA purification (step 67).
PAUSE STEP Either proceed with the protocol or flash-freeze the pellets of nuclei in liquid nitrogen (or dry ice/ethanol bath) and store them at -80 °C.
. PAUSE STEP Either proceed with the protocol or flash-freeze the cell pellets in liquid nitrogen (or a dry ice/ethanol bath) and store them at -80 • C. We have compared both approaches, and freezing the cells did not affect the efficiency of subsequent steps. 3. Count the cells (for example, using a hemocytometer or an automated cell counter) and use 10 million cells per HiChIP experiment or 5 million cells per Hi-C experiment. Bring up the volume to 10 mL with 1× PBS and resuspend the cells well. 10. Repeat the wash (steps 8-9) one more time and discard the supernatant.

Lysis and Restriction
. PAUSE STEP Either proceed with the protocol or flash-freeze the cell pellets in liquid nitrogen (or a dry ice/ethanol bath) and store them at -80 °C. We have compared both approaches, and freezing the cells did not affect the efficiency of subsequent steps. . CRITICAL STEP Efficient preparation of nuclei is critical, so we advise monitoring cell lysis under a microscope upon Trypan blue staining, which should not stain the cells with an intact cellular membrane. Extended cellular lysis up to 4 h did not negatively affect the experiment, unlike the insufficient lysis.

Lysis and Restriction Enzyme Digestion (Required
12. Spin down the nuclei at 2500× g at 4 °C for 5 min and discard the supernatant. 13. Wash the pellet once with 500 μL of ice-cold Hi-C Lysis Buffer. . CRITICAL STEP We noticed that older batches of Triton X-100 tend to have low pH, which negatively affects the enzymatic reactions. . CRITICAL STEP Efficient preparation of nuclei is critical, so we advise monitoring cell lysis under a microscope upon Trypan blue staining, which should not stain the cells with an intact cellular membrane. Extended cellular lysis up to 4 h did not negatively affect the experiment, unlike the insufficient lysis. 10. Repeat the wash (steps 8-9) one more time and discard the supernatant.
. PAUSE STEP Either proceed with the protocol or flash-freeze the cell pellets in liquid nitrogen (or a dry ice/ethanol bath) and store them at -80 °C. We have compared both approaches, and freezing the cells did not affect the efficiency of subsequent steps.
. CRITICAL STEP Efficient preparation of nuclei is critical, so we advise monitoring cell lysis under a microscope upon Trypan blue staining, which should not stain the cells with an intact cellular membrane. Extended cellular lysis up to 4 h did not negatively affect the experiment, unlike the insufficient lysis.
12. Spin down the nuclei at 2500× g at 4 °C for 5 min and discard the supernatant. 13. Wash the pellet once with 500 μL of ice-cold Hi-C Lysis Buffer.
14. Remove the supernatant and gently resuspend the pellet by pipetting up and down in 100 μL of 0.5% SDS.
15. Incubate at 62 °C for 10 min and then add 300 μL of ddH2O and 50 μL of 20% Triton X-100 to quench the SDS. 16. Mix by pipetting and incubate at 37 °C for 15 min.
17. Add 50 μL of 10× DpnII Buffer and check the pH with 4 μL spotted on a pH-indicator paper. If it is in the range of pH 6-7, you can proceed; otherwise, fix the pH with the addition of Tris-HCl pH 8.0.
. CRITICAL STEP We noticed that older batches of Triton X-100 tend to have low pH, which negatively affects the enzymatic reactions. . CRITICAL STEP We noticed that older batches of Triton X-100 tend to have low pH, which negatively affects the enzymatic reactions.   Table 1 below.
Merge the two halves of each HiChIP sample back together (does not apply to Hi-C), mix it carefully by pipetting and keep 15 µL as the control C1.
Incubate at 37 • C for 5 min.

28.
Split the HiChIP samples into two halves again, before ligation (do not split the Hi-C sample).

29.
Spin down the nuclei at 2500× g at 4 • C for 10 min and discard the supernatant. 30.
Resuspend the nuclei in 1200 µL of Ligation Master Mix according to Table 2 below. mix carefully by pipetting and take 120 µL as the control C2. 33. Spin down the nuclei at 2400× g at 4 • C for 15 min and discard the supernatant. 34. To further decrease the number of dangling ends, we recommend implementing the removal of unligated, biotinylated ends [17]. For HiChIP libraries, it can either be performed right after the ligation in T4 ligation buffer (step 32) or after nuclei spin down (at this step, in its dedicated T4 ligase reaction buffer). For the reaction, add only dGTP and dATP nucleotides and T4 DNA polymerase, which will naturally manifest its 3 -5 exonuclease activity and remove the unligated 3 biotin-dCTP. Alternatively, this step can be performed after DNA purification (step 67). 28. Split the HiChIP samples into two halves again, before ligation (do not split the Hi-C sample).
29. Spin down the nuclei at 2500× g at 4 °C for 10 min and discard the supernatant. 30. Resuspend the nuclei in 1200 μL of Ligation Master Mix according to Table 2 below. 34. To further decrease the number of dangling ends, we recommend implementing the removal of unligated, biotinylated ends [17]. For HiChIP libraries, it can either be performed right after the ligation in T4 ligation buffer (step 32) or after nuclei spin down (at this step, in its dedicated T4 ligase reaction buffer). For the reaction, add only dGTP and dATP nucleotides and T4 DNA polymerase, which will naturally manifest its 3′-5′ exonuclease activity and remove the unligated 3′ biotin-dCTP. Alternatively, this step can be performed after DNA purification (step 67).

PAUSE STEP
Either proceed with the protocol or flash-freeze the pellets of nuclei in liquid nitrogen (or dry ice/ethanol bath) and store them at -80 °C.
PAUSE STEP Either proceed with the protocol or flash-freeze the pellets of nuclei in liquid nitrogen (or dry ice/ethanol bath) and store them at -80 • C. . PAUSE STEP Either proceed with the protocol or flash-freeze the cell pellets in liquid nitrogen (or a dry ice/ethanol bath) and store them at -80 °C. We have compared both approaches, and freezing the cells did not affect the efficiency of subsequent steps. . CRITICAL STEP Efficient preparation of nuclei is critical, so we advise monitoring cell lysis under a microscope upon Trypan blue staining, which should not stain the cells with an intact cellular membrane. Extended cellular lysis up to 4 h did not negatively affect the experiment, unlike the insufficient lysis.

Lysis and Restriction Enzyme Digestion (Required
12. Spin down the nuclei at 2500× g at 4 °C for 5 min and discard the supernatant. 13. Wash the pellet once with 500 μL of ice-cold Hi-C Lysis Buffer.
14. Remove the supernatant and gently resuspend the pellet by pipetting up and down in 100 μL of 0.5% SDS.
CRITICAL STEP SDS will start precipitating when on ice, so try to avoid it by controlling the sample and putting it on and off the ice.  Figure 2c). 68. Additionally, you can use a part of the C3 control sample to set up a control PCR reaction. Ligation of two DpnII blunt DNA ends produces a novel restriction site recognized by the restriction enzyme ClaI [17]. Hence primers designed for two adjacent restriction fragments with a high likelihood to be ligated together can be used to amplify the ligated molecule. If the fragments were properly digested, filled in and re-ligated, then they should be digested by ClaI. The ratio between the digested and undigested fragments provides another piece of evidence about the protocol's efficiency.
29. Spin down the nuclei at 2500× g at 4 °C for 10 min and discard the supernatant. 30. Resuspend the nuclei in 1200 μL of Ligation Master Mix according to Table 2 below. 34. To further decrease the number of dangling ends, we recommend implementing the removal of unligated, biotinylated ends [17]. For HiChIP libraries, it can either be performed right after the ligation in T4 ligation buffer (step 32) or after nuclei spin down (at this step, in its dedicated T4 ligase reaction buffer). For the reaction, add only dGTP and dATP nucleotides and T4 DNA polymerase, which will naturally manifest its 3′-5′ exonuclease activity and remove the unligated 3′ biotin-dCTP. Alternatively, this step can be performed after DNA purification (step 67).
PAUSE STEP Either proceed with the protocol or flash-freeze the pellets of nuclei in liquid nitrogen (or dry ice/ethanol bath) and store them at -80 °C. 85. Resuspend the beads in 50 µL of PCR master mix prepared according to Table 3. Due to different Nextera barcodes for each sample, it needs to be pipetted individually.  Table 4 with the number of repeated cycles estimated based on the amount of post-ChIP DNA determined by Qubit. An accurate estimation of the number of PCR cycles is never an easy task. One option is to follow the recommendation from the original HiChIP paper: "Greater amount than 50 ng run for 5 cycles, approximately 50 ng for 6 cycles, 25 ng for 7 cycles, 12.5 ng for 8 cycles, etc" [9]. However, the quality of libraries is not reflected in the quantification, so this approach is really just an estimation (although often sufficient). In our laboratory, we tend to amplify a bit more, ranging between 8-14 cycles for 50-3 ng of post-ChIP DNA, respectively, in order to obtain a sufficient amount of DNA.

Expected Results
Despite all the available controls, we recommend running a pilot low-scale sequencing experiment to evaluate the quality of libraries. In Figure 4a, we demonstrate that even as few as 2.5 million sequencing reads are enough to obtain a picture of chromosome architecture. Moreover, this small sequencing experiment provided almost identical proportions of valid interaction pairs when compared to the main sequencing experiment; thus, it can be used as a valid estimation of the library quality.
In our protocol, we usually detect a high percentage of inter-chromosomal interaction pairs (Figure 4b). In the original protocols performing the ligation step in solution, these were deemed a sign of a low-quality library as they originate from random chimeric formation between DNA fragments originating from different cells [29]. The approach used in our protocol, performing the ligation step in the nucleus, ensures the high specificity of inter-chromosomal interactions. Moreover, the frequent centrifugation steps and washes ensure that broken nuclei potentially contributing to false chimeric trans interactions would be washed away. The high percentage of inter-chromosomal interactions is likely a result of the enhanced ligation efficiency using PEG and partially due to the default compactness of the T cell nuclei. Our group has previously demonstrated the biological significance of inter-chromosomal encounters [30]; hence, it is desired not to exclude them from the analysis. The relative proportion of short and long cis and trans chromatin interactions, compared to other published studies, are depicted in Figure 4c. Apart from the high number of inter-chromosomal interactions, we would like to highlight the limited number of short-range interaction pairs (<20 kbp), which emphasizes the high efficiency of the protocol. This protocol was successfully used to investigate the 3D enhancer network of murine thymocytes using SATB1 and CTCF HiChIP experiments in wild-type cells and Hi-C and H3K27ac HiChIP experiments in both wild-type and Satb1 fl/fl Cd4-Cre + cells [31].