Automated ChIPmentation procedure on limited biological material of the human blood fluke Schistosoma mansoni

Automated ChIPmentation procedure is a convenient alternative to native chromatin immunoprecipitation (N-ChIP). It is now routinely used for ChIP-Seq. Using the human parasite Schistosoma mansoni, whose production requires scarifying animals and should therefore kept to a minimum, we show here that the automated ChIPmentation is suitable for limited biological material. We define as operational limit ≥20,000 cells. We also present a streamlined protocol for the preparation of ChIP input libraries.


Introduction
Schistosoma mansoni is a human parasite with a complex life cycle that shows strong developmental phenotypic plasticity, with intra-molluscal, and intra-vertebrate stages and two free-swimming larvae stages (miracidium and cercariae). We had shown by native chromatin immunoprecipitation (N-ChIP) that the different life cycle stages show also strong histone modification plasticity (Cosseau et al. 2009, Roquis et al. 2018, de Carvalho Augusto et al. 2020. While N-ChIP is in principle doable we found that there are two challenges associated with: one is the high hands-on time with the N-ChIP and the other is to obtain enough biological material for performing several ChIP experiments with different antibodies. We therefore explored an automated ChIP procedure ( Figure 1) and we tested what the lowest reasonable cell number would be with this procedure.
ChIPmentation is a ChIP-sequencing (ChIP-seq) technology which uses a transposase to add the sequencing adaptors to the DNA of interest instead of the classical multi-step processing including end repair, A-tailing, adaptor ligation and size-selection (Schmidl et al. 2015). Thanks to the action of the transposase, loaded with sequencing adaptors, the library preparation is performed in only one step, which reduces hands-on time and material loss. Moreover, in the ChIPmentation approach this tagmentation process is performed directly on chromatin during the immunoprecipitation process, instead of naked DNA after purification. This workflow allows for a more reproducible tagmentation. for the fragmentation step. The immunoprecipitation is done by centrifugation. All the process is manually done.
ChIPmentation protocol lasts 2 days. Cross-linking is used for cell lysis and sonication for the fragmentation step. The immunoprecipitation, tagmentation and library cleaning are done with the IP-Star.
. CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted June 2, 2021. ; https://doi. org/10.1101org/10. /2021 Page 4 of 18 The combined facts that ChIPmentation has been automated on Diagenode's IP-Star Compact Automated System and that this technology has been validated on low amounts of human cells (Schmidl et al. 2015, Roels et al. 2020) make it a perfect candidate for ChIP-seq of various targets on limited Schistosoma mansoni material.

Results
ChIPmentation has a sensitivity that is comparable to N-ChIP 10 6 -10 4 cell equivalents gave comparable Bioanalyzer formats with peaks around 1kb, from 5,000 cells equivalent on fragments of smaller size became clearly visible. No high molecular fragments were observed in the negative control without chromatin (figure 2, table 1). L1-8 R1-8 . CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted June 2, 2021. ; https://doi. org/10.1101org/10. /2021 Alignment gave expected results (~50% uniquely aligned reads) for 10 6 -10 4 cell equivalent but dropped to ~35% with 10 3 cells equivalent and <20% for 100 cells equivalent. No contaminating DNA was detected (table   2).  . CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted June 2, 2021. ; https://doi.org/10. 1101/2021 In order to compare ChIPmentation results to Native-ChIP (N-ChIP) we re-analysed earlier data obtained by N-ChIP (Augusto, Cosseau et al. 2019) using the same data cleaning, alignment and peak calling parameters as for ChIPmentation (table 3).
For ChIPmentation, peakcalling with Peakranger was very robust for 10 6 cells and delivered the expected values (based on earlier N-ChIP results). Below this cell equivalent, peak calling became dependent on bin size (   selected as the best extension length and applied to N-ChIP data below. N-ChIP A is for 0.8x10 6 cells. . CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted June 2, 2021. ; https://doi.org/10.1101/2021.06.02.446743 doi: bioRxiv preprint We iteratively identified 1,000 bp as best read extension length. The use of HMM-based ChromstaR improved peak calling for ChIPmentation 10 4 cell equivalent, but not for 10 3 or 100 cells. We obtained comparable results for ChIPmentation of 10 6 , 10 4 cells and N-ChIP ~10 6 , 10 4 and 10 3 cells (table 5).  ChromstaR metagene profiles showed consistent profiles for ChIPmentation 10 6 , 10 4 and 10 3 cells, and all N-ChIP, but not for ChIPmentation on 100 cells equivalents (figure 3). . CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted June 2, 2021. ; https://doi.org/10.1101/2021.06.02.446743 doi: bioRxiv preprint ChromstaR allows for estimating correlation of chromatin profiles based on read counts (figure 4), and indicates high correlation between ChIPmentation 10 6 , and N-ChIP ~10 6 , 10 4 .
All other over-all chromatin profiles were below 0.75 correlation coefficient. This is surprising given the high similarity of metagene profiles. Visual inspection of peaks and profiles showed that peaks were actually correctly identified by ChromstaR (but much less by Peakranger) in ChIPmentation until 10 4 cells but there is higher background than in N-ChIP which probably spoils correlation for lower cell equivalents (figure 5). . CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted June 2, 2021. ; https://doi.org/10.1101/2021.06.02.446743 doi: bioRxiv preprint ChIPmentation input library can rapidly be produced in parallel to the automated procedure After having formally established that automated ChIPmentation had a comparable sensitivity to our routine N-ChIP procedure we aimed to identify the optimal way to produce input libraries for control of unspecific enrichment. The production of input chromatin is "build-in" the N-ChIP protocol (Cosseau et al. 2009, Roquis et al. 2018, de Carvalho Augusto et al. 2020) and need to be adapted to the automated ChIPmentation procedure. During a ChIPmentation experiment three types of input can be imagined (figure 6): i-1µL of chromatin before immunoprecipitation, ii-chromatin that binds non-specifically to any support and iiiavailable chromatin for immunoprecipitation. The aliquot of i-1µL is taken from the sample before immunoprecipitation. The two other types (input library ii and iii) need a supplementary sample in which mock . CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted June 2, 2021. ; https://doi.org/10.1101/2021.06.02.446743 doi: bioRxiv preprint immunoprecipitation is done without antibody. After immunoprecipitation, this supplementary sample contains magnetic beads with the non-specifically bound chromatin and the supernatant which is the available chromatin for immunoprecipitation. Figure 6: Type of inputs generated during ChIPmentation and its bioanalyser profiles associated. X axis represents the size in base pair, and Y axis represents the fluorescence intensity. Ideal library size is between 150 bp and 500 bp. i-1µL aliquot of one sample chromatin took before its immunoprecipitation. ii-and iii-are from a supplementary sample where immunoprecipitation in the IP-Star is done without antibody. After immunoprecipitation, this supplementary sample contains magnetic beads with the iinon-specifically binding chromatin and the supernatant which is the iii-available chromatin for immunoprecipitation. ii-No specific binding chromatin gave not usable library.
. CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted June 2, 2021. ; https://doi.org/10. 1101/2021 Only i-before immunoprecipitation chromatin and iii-available chromatin for immunoprecipitation give ideal library sizes (figure 6). We decided to optimize the input protocol for the i-1µL of non-immunoprecipitated chromatin because it does not occupy a slot in the IP-Star. In addition, during the preliminary test, 1µL chromatin input showed a smaller Ct compare to the available chromatin input which means that it requires less amplification cycles (data not shown).
For the preparation of the 1µL chromatin input libraries, we identified two critical parameters: the first one is the dilution of the tagmentation enzyme. Using undiluted Tn5 caused complete overtagmentation. Between 10-fold and 100-fold dilutions in water delivered much better results ( figure 7-A).
. CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted June 2, 2021. ; https://doi.org/10.1101/2021.06.02.446743 doi: bioRxiv preprint

Discussion
Here we operationally define the limit for the robust detection of peaks with ChIPmentation to be 100,000 cells equivalent per antibody reaction, with 10,000 being the absolute limit if background is acceptable.
Operational limit for N-ChIP is 10,000 cells equivalent, confirming our previous results. We therefore established ≥20,000 cells as lower limit for the routine ChIPmentation-seq procedures in S. mansoni. This is a bit higher than what has been described in Human cells where good results have been obtained on as low as 5,000 cells, but this is consistent with the fact that the genomic content of one cell is not the same between the two species.
To improve signal to noise ratio and reduce background in ChIPmentation it could be useful to increase the washing time (currently 10 min) and speed (currently medium) but we recommend increasing cell number than to invest into washing optimization.
1µL non-immunoprecipitated chromatin input is the best compromise to save space in the IP-Star, experiment time and biological materials when you are restricted by quantity. This ChIPmentation protocol is not limited to miracidia cells. We also performed this protocol on S. mansoni adult worms and snail hepatopancreas infected by diurnal and nocturnal S. mansoni from Oman sporocysts (data not published). Remember that before to do any ChIPmentation experiment, it is necessary to determine the amount of antibody and the sonication time adapted to your model.

Cell lysis and chromatin shearing
Chromatin preparation was performed using the Diagenode's ChIPmentation Kit for Histones, Cat. No.
C01011009 and protocol with minor modifications. Two times 10,000 miracidia (1,000,000 cells based on the observation that 1 miracidium is composed of 100-120 cells) were resuspended each in 500 µL 1x HBSS and crashed with a plastic pestle in an Eppendorf tube on ice during ~1 min. For crosslinking, 13.5 µL of formaldehyde were added and tubes were incubated for 10 min at room temperature with occasional inversion.
To stop crosslinking fixation, 57 µL of glycine were added and samples were incubated for 5 min at room temperature. Samples were centrifuged at 500xg, at 4°C, for 5 min. The pellet was resuspended in 2x1 mL of ice-cold Lysis Buffer iL1, combined and homogenized in a Dounce (pestle A) on ice for 5 min. After another centrifugation (500g, 5 min, 4°C), the supernatant was discarded and the pellet was resuspended in 1 mL of ice-cold Lysis Buffer IL2 and centrifuged (500xg, 5 min, 4°C). The supernatant was discarded and the pellet was resuspended in 100 µL of complete Shearing Buffer iS1 for each tube. Samples were sonicated with the Bioruptor Pico (Diagenode, Cat. No. B01080010 ) for 5 cycles (30s ON and 30s OFF). After transfer into new tubes, samples were centrifuged (16,000g, 10 min, 4°C). The supernatants were transferred into a new single tube (200 µl total) and 20 µl iS1 were added yielding 220 µl total volumes. The procedure was done in duplicates (named L and R in the following). Serial dilutions were done in iS1 to produce 100 µl equivalents of 10,000 miraciadia (10 6 cells), 1,000 miracidia (10 5 cells), 100 miracidia (10 4 cells), 50 miracidia (5,000 . CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted June 2, 2021. ; https://doi.org/10.1101/2021.06.02.446743 doi: bioRxiv preprint cells), 10 miracidia (10 3 cells), 5 miracidia (500 cells) and 1 miracidium (100 cells) or 100 µl iS1 as negative control. Stripping, end repair and reverse cross-linking were done as indicated in the User Guide.

Input library tagmentation
In the ChIPmentation Kit for Histones (Diagenode, Cat. No. C01011009) the suggested strategy is to sequence one sample immunoprecipitated with a control IgG and to use it for sequencing normalization instead of the classical input which cannot be treated exactly the same way as the immunoprecipitated samples. But IgG are negative control samples and the generation of such samples on low amounts approaches involves the use of a very high number of amplification cycles that can induce some biases. A protocol for the tagmentation of the input sample has therefore been set-up as follows.
For each immunoprecipitated sample, 1µL of sheared chromatin was kept aside before IP in the IP-Star. 1µL of MgCl2 (Diagenode ChIPmentation kit for Histones, Cat. No. C01011009), 8µL of molecular biology grade water, 10µL 2xTagment DNA buffer and 1µL of 100-fold in molecular-grade water diluted DNA tagmentation enzyme (Illumina 20034197, lot 20464427) were added to each 1µL input. The tagmentation reaction was performed in a thermocycler for 5 minutes at 55°C. Then, 25µL of 2xPCR NEB master mix (New England Biolabs M0541L, lot 10067165) was added to each input. The end-repair and de-cross-link were performed in a thermocycler for 5 minutes at 72°C followed by 10 minutes at 95°C. An aliquot of 2µL was taken from each input and added to 8µL of quantification mix. For each reaction, this quantification mix is composed of 0.3 µL of forward and reverse ATAC-seq primers (25µM) (suppl. table 1, Buenrostro et al. 2015), 1µL SYBR green 10X (Diagenode kit), 1.3 µL of molecular biology grade water and 5µL of 2xPCR NEB master mix. While not . CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted June 2, 2021. ; https://doi.org/10.1101/2021.06.02.446743 doi: bioRxiv preprint formally tested, leftover primers of the ChIPmentation kit could probably also be used but volume must be adjusted as the primer pairs in the kit are at 10 µM instead of 25 µM. Library amplification, purification, quality checking and sequencing steps were performed as for the immunoprecipitated samples (see below).

Library amplification
To determine optimal number of library amplification cycles we proceeded as described in step 5.1 to 5.5 of ChIPmentation User Guide with minor modifications. We determined the number of amplification cycles for each library by using the number of cycles that corresponded to 1/3 of the slope of qPCR amplification curve during the exponential phase. Results are in table 1 Amplification was done in step 5.7. After PCR, 48 µl of library amplification mix were AMPure purified on the IP-Star using 86 µl of AMPure beads (1.8x). Instead of resuspension buffer we used ChIP grade water.
After that libraries were in 20 µl.

Library check and Sequencing
Library fragment size and concentration was checked on an Agilent 2100 Bioanalyzer with a High Sensitivity DNA Assay v1.03. Paired-end sequencing (2x75 cycles) was performed at the Bio-Environnement platform (University of Perpignan, France) on a NextSeq 550 instrument (Illumina, USA).

Bioinformatics Analysis
Reads were quality checked with FastQC. Adapters were detected in less than 4% of reads. Reads were aligned to the S. mansoni v7 reference genomes using Bowtie2, sensitive settings and default values.
Uniquely aligned reads were filtered from the BAM files using the XS: tag of Bowtie. PCR duplicates were removed with samtools rmdup. BAM files were subsampled to 3.8 Mio uniquely aligned reads with picard DownsampleSam. Peakcalling was done with Peakranger P value cut off: 0.0001, FDR cut off: 0.05, Read extension length 100 -2,000 bp, Smoothing bandwidth: 99. Delta: 0.8, Detection mode: region. ChromstaR was used with a bin size same as Peakranger extension lengths and a step size of half of bin size. Miracidia genomic DNA libraries served as input. In ChromstaR postprocessing, maximum posterior probabilities to . CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted June 2, 2021. ; https://doi.org/10.1101/2021.06.02.446743 doi: bioRxiv preprint adjust sensitivity of peak detection was set to 0.999. This keeps broad peaks intact. We reasoned that gaps between peaks that were larger than a nucleosome are not biologically meaning full and peaks were merged with BEDTOOLS when they were ≤150 bp apart (the average length of DNA in a nucleosome).
Enrichment plots over metagenes were produced over 5,073 genes on the +strand based on canonical gff v7 of S. mansoni.

Data availability
Fastq files are available at the NCBI SRA Project accession number XXXX. . CC-BY-NC-ND 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted June 2, 2021. ; https://doi.org/10.1101/2021.06.02.446743 doi: bioRxiv preprint