Cell-free Chromatin Immunoprecipitation to detect molecular pathways in Physiological and Disease States

Patient monitoring is a cornerstone in clinical practice to define disease phenotypes and guide clinical management. Unfortunately, this is often reliant on invasive and/or less sensitive methods that do not provide deep phenotype assessments of disease state to guide treatment. This paper examined plasma cell-free DNA chromatin immunoprecipitation sequencing (cfChIP-seq) to define molecular gene sets in physiological and heart transplant patients taking immunosuppression medications. We show cfChIP-seq reliably detect gene signals that correlate with gene expression. In healthy controls and in heart transplant patients, cfChIP-seq reliably detected housekeeping genes. cfChIP-seq identified differential gene signals of the relevant immune and non-immune molecular pathways that were predominantly downregulated in immunosuppressed heart transplant patients compared to healthy controls. cfChIP-seq also identified tissue sources of cfDNA, detecting greater cell-free DNA from cardiac, hematopoietic, and other non-hematopoietic tissues such as the pulmonary, digestive, and neurological tissues in transplant patients than healthy controls. cfChIP-seq gene signals were reproducible between patient populations and blood collection methods. cfChIP-seq may therefore be a reliable approach to provide dynamic assessments of molecular pathways and tissue injury associated to disease.


1
Defining and monitoring the molecular phenotype of disease guides clinical decision-making 2 including treatment selection, assessment of disease progression and response to treatment. 3 Unfortunately, for many conditions, invasive approaches such as tissue biopsy remain the gold 4 standard. Heart transplant patients, for example, undergo 15 -25 endomyocardial biopsies to 5 monitor for acute rejection in the first year of transplantation alone. In addition to a high cost 6 and risk of complications, biopsy samples are examined by histopathology, which is limited by 7 low sensitivity and high inter-operator variability (Marboe et al. 2005). Emerging blood-based 8 approaches may address these limitations (Moss et al. 2018;Andargie et al. 2021;Sadeh et al. 9 2021; Vorperian et al. 2022). Recently, cell-free DNA (cfDNA) chromatin immunoprecipitation 10 sequencing (cfChIP-seq) has been proposed as one of such an approach; being minimally invasive 11 and reliable to define molecular phenotypes and the tissue types involved in multiple disease 12 states (Sadeh et al. 2021). Such an approach could serve as a significant advancement to monitor 13 allograft health. Like other sequencing-based approaches, reproducibility and standardization is 14 important to enable broad applicability and interpretation of cfChIP-Seq results across studies. 15 Circulating cfDNA, released into the bloodstream following cell death, is attracting great 16 attention as a novel biomarker for early diagnosis and monitoring in a range of disease conditions 17 (Agbor-Enoh et al. 2019;Duvvuri and Lood 2019;Jackson Chornenki et al. 2019;Zviran et al. 2020;18 organs, is a reliable alternative to biopsy. Current approaches use donor-recipient single 1 nucleotide polymorphisms to quantitate donor-derived cfDNA as measure of allograft injury. 2 Although the SNP-based approach is sensitive to detect allograft rejection and other 3 complications, it lacks specificity to identify acute rejection phenotypes or define the molecular 4 pathways involved to tailor treatment. CfDNA is histone bound and maintains histone 5 modifications from its tissue of origin (Sadeh et al. 2021). Post-translational modifications of 6 histones can regulate genomic elements and is often a proxy for gene expression. The associated 7 genes may show cell/tissue specificity, identifying tissue source involved in disease (Sadeh et al. 8 2021). Thus, cfChip-seq can delineate different disease processes, annotate disease phenotypes, 9 and identify tissue injury patterns (Sadeh et al. 2021). 10 Despite advances in the development of novel and specific diagnostic approaches, reproducibility 11 remains a major limitation for genomic approaches. In a review of sequence-based studies, only 12 ~25% of the studies provide sufficient information to enable adequate technical and biological 13 reproducibility (Nekrutenko and Taylor 2012). For cfDNA-based studies, blood collection is an 14 added matrix of variability with different collection protocols. Standardization is important to 15 account for and/or limit contamination of cfDNA with cellular genomic DNA from cell lysis during 16 plasma preparation. In one heart transplant study, most of the plasma samples collected were 17 not analyzable because of cfDNA quality, even though specialized blood collection tubes with 18 preservative to prevent blood cell lysis were used (Richmond et al. 2020). Published cfDNA 19 studies continue to use tubes with different preservatives to collect blood and processed blood 20 at different centrifugation speed (Lo et al. 1999;Agbor-Enoh et al. 2017;Sadeh et al. 2021). It 21 remains unknown how these differences contribute to cfDNA results.

22
. CC-BY 4.0 International license 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 made The copyright holder for this preprint this version posted January 25, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 Here, in addition to assessing its reproducibility, we assessed if cfChIP-seq can be used to define 1 biological processes in immunosuppressed transplant patients compared to healthy controls, as 2 a first step towards developing specific nucleosome-based cfDNA test that can define molecular 3 states of disease. 4 5 . CC-BY 4.0 International license 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 made The copyright holder for this preprint this version posted January 25, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 Results 1 This study intends to assess if a cfChIP-seq approach would delineate biological processes in 2 physiological conditions and, in heart transplantation, as a prototype of disease state. Further, 3 the study assesses the reproducibility of cfChIP-seq between different blood collection methods. 4 Housekeeping genes in healthy controls were selected to reflect physiological conditions. 5 Differential gene signals between immunosuppressed heart transplant patients and healthy 6 controls were performed as a prototype of defined molecular pathways in a disease state. The 7 study involved collection of blood samples and performance of cfChIP-seq on 8 healthy control 8 plasma samples and 6 samples from 2 heart transplant recipients (2 prior to transplantation and 9 4 post-transplantation). Subjects gave consent. To assess if blood collection methods affect 10 cfChIP-seq results for healthy controls, paired blood samples were collected into two different 11 blood collection tubes; Streck tubes, selected as a prototype sample collection tube containing a 12 proprietary preservative that prevents cell lysis, and EDTA tubes without any anti-cell-lysis 13 preservative. Plasma was separated from blood cells within two hours of blood collection by 14 centrifugation at 1,600 g, followed by centrifugation at two different speeds (3,000 g vs. 16,000 15 g). So, in total, replicate samples were processed under four conditions: (1) Streck tubes spun at 16 16,000g (S16), (2) Streck tubes spun at 3000g (S3), (3) EDTA tube spun at 16,000g (E16) and (4)   17 EDTA tubes centrifuged at 3000g (E3) (Figure 1a). 18 For cfChip-seq, we adopted the protocol from Sadeh et al. (Sadeh et al. 2021). In summary, 19 antibodies directed to specific histone modifications were coupled to magnetic beads and 20 incubated with 1 mL of thawed plasma. After washing and digestion of bound histones using 21 . CC-BY 4.0 International license 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 made The copyright holder for this preprint this version posted January 25, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 proteinase K, captured cfDNA were indexed using Accel-NGS 2S plus DNA library kit with Unique 1 Dual Indexing (Swift Biosciences). The cfDNA libraries were subject to paired-end sequencing on 2 the NovoSeq platform at ~140 million reads per sample. Two controls were included for each 3 sample, non-specific IgG and input cfDNA. Length distribution of sequence reads showed 4 nucleosomal distribution with mononucleosomal predominance as expected (Figure 1b). Of the 5 four histone antibodies, H3K4me2 and H3K4me3 reached saturation at ~30,000 million read pairs 6 per sample; the two other histone antibodies (H3K4me1 and H3K36me3) and input cfDNA did 7 not reach saturation (Figure 1c, Suppl. Figure 1a). The three H3K4me marks followed the same 8 global distribution (Figure 1d). Their local distributions matched known patterns, with H3K4me3, 9 for example, found mostly near transcriptional start sites (TSS) (Figure 1e), H3K4me2 found at 10 enhancers and TSS sites, and H3K4me1 found primarily at enhancers of housekeeping (active) 11 genes. The number and called peaks for each histone were consistent with these patterns. For 12 example, 46% of H3K4me3 peaks overlapped promoters, compared to 15% for K4Me2 and 18% 13 for K4Me1 (Suppl. Table 1). Use of the input control samples, increased the power of our 14 analyses. It increased the number of peaks identified for all histone modifications (Supp. Table   15 1). It also corrected for a weak binding pattern seen on gene bodies for non-expressed genes for 16 all histones (Figure 1e, Suppl. Fig 1b). For these reasons, input-normalization was utilized for all 17 downstream analyses.

18
Reproducibility of cfChIP-seq signals across blood collection methods 19 We first evaluated if the four different blood collection and processing conditions alter cfChIP 20 gene signals for H3K4me3, H3K4me1, H3K4me2, and H3K36me3. All four conditions, i.e S3, S16, 21 E3, and E16, showed expected peak frequency distribution for H3K4me3 around the TSS for 22 . CC-BY 4.0 International license 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 made The copyright holder for this preprint this version posted January 25, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 expressed genes consistent with known promoter location (Figure 2a). Expected frequency 1 distribution of gene signals was also observed for H3K4me1, H3K4me2, and H3K36me3 (Suppl. 2 Figure 2). For all four blood processing conditions, FRiP were highest for H3K4me3 than for 3 H3K4me1 and H3K4me2 (Figure 2b). Fraction of reads in peaks (FRiP) was lower and/or more 4 variable between subjects for E16 and E3 than for S16 and S3; E3 showing the most variable FRiP, 5 while S3 and S16 showed more consistent FRiP between subjects for H3K4me1, H3K4me2, and 6 H3K4me3 (Figure 2b). The decrease in FRiP in EDTA tubes compared to Streck tubes was 7 significant (p-value: 2e -4 ). The number of peaks detected was equally variable for E3, E16 than 8 for S3 and S16 (Figure 2c). Despite the variability in peaks between the conditions, for all four 9 conditions, H3K4me3 cfChIP-seq showed similar fraction of promoters overlapping peaks ( Figure   10 2d), as well as a similar number of genes detected (Suppl. Table 1). H3K4me1 and H3K4me2 also 11 showed similar number of genes detected, except for one sample that showed lower number of 12 genes detected for E3 (Suppl. Table 1).

13
CfChIP-seq reliably detect housekeeping genes 14 We next assessed if cfChIP-seq gene signals are reflective of gene expression, focusing on 15 S16/H3K4me3, which showed highest and most consistent RPGC signal between subjects. 16 Prototype housekeeping genes (GAPDH and TBP) show peaks matching their promoter location, 17 around the TSS. Similarly, example monocyte-specific genes (FCN1 and CSF3R), a third major cell 18 type contributing plasma cfDNA, showed peaks matching their gene promoter location. However, 19 the monocyte-specific genes showed lower RPGC compared to housekeeping genes that are 20 constitutionally expressed in all tissues. Non-expressed genes in healthy patients (IL-3 and CSF2) 21 showed no or non-specific peaks, with baseline levels that are no different from non-specific IgG 22 . CC-BY 4.0 International license 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 made The copyright holder for this preprint this version posted January 25, 2023. ; https://doi.org/10.1101/2023.01.24.525414 doi: bioRxiv preprint (Figure 3a). In total, H3K4me3 cfChIP-seq detected 93% of housekeeping genes, (Figure 3b), that 1 is 8486 of the 9099 housekeeping genes represented in Suppl. Table 2. Of the 4216 non-2 housekeeping genes detected by H3K4me3 cfChIP-seq, one-third (n=1475 genes) were 3 neutrophil and/or monocyte-specific genes; neutrophils and monocytes contribute over one-  CfChIP-seq identifies relevant molecular pathways in heart transplant patients 11 We next analyzed plasma from heart transplant recipients maintained on immunosuppression 12 drugs. The patients involved all reached theurapeutic tacrolimus blood toughs by 2 weeks of 13 transplantation. So, post-transplant blood samples were collected 28 -318 days after 14 transplantation. Blood samples were collected in Streck tubes and spun to 1,600g followed by 15 16,000g. The isolated cfDNA showed an expected nucleosomal pattern (Suppl. Figure 3a). 16 H3K4me3 peaks showed highest frequency around TSS as expected (Suppl. CC-BY 4.0 International license 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 made The copyright holder for this preprint this version posted January 25, 2023. ; https://doi.org/10.1101/2023.01.24.525414 doi: bioRxiv preprint signals and not due to different in the healthy control and heart transplant groups, we included 1 two pre-transplant plasma samples collected from the two heart transplant recipients. Both 2 heart transplant recipients had advanced dilated cardiomyopathy and were status 1A on the 3 transplant wait list at the time of sample collection. PCA showed separation of cfChIP-seq signals 4 for the pre-transplant samples versus healthy controls, as well as separation of signals for the 5 pre-transplant and post-transplant samples (Figure 4b). 6 We employed the Deseq2 version of DiffBind v2 (Ross-Innes et al. 2012) to identify differential 7 peaks between healthy controls and heart transplant recipients. There were 4030 peaks with a 8 FDR less than 0.05 between the healthy controls and heart transplant samples (Figure 4c). 48.8% 9 of these peaks overlapped promoters of 2137 genes. Pathway analysis of these genes identified 10 immune and non-immune pathways that are relevant to transplantation (Figure 4d, Suppl. Table   11 3). The immune pathways were predominantly associated with a loss of H3K4Me3 in their 12 promoters in transplant recipients versus healthy controls, correlating with the Table 3). Genes associated to re-organization of extracellular matrix and fibrosis showed 1 differences between transplant and healthy controls, including genes associated to 2 glycosaminoglycan metabolism (n=26), extracellular matrix organization (n=57), collagen 3 synthesis (n=21), and integrin synthesis/re-organization (Suppl . Table 3). Interesting, multiple 4 gene sets associated to neuropathy were also differentially detected in transplant patients, 5 neuropathy due to drug toxicity is a common manifestation in heart transplant patients.  (Figure 5b). Comparing the 10 four blood processing methods, S16 and S3 produced more consistent cfDNA tissue contributions 11 than E16 or E3 (Suppl. Fig 4a). Vascular endothelial cells, pulmonary tissues and other non-12 hematopoietic tissue types were also detected. The fraction of tissue-specific cfDNA detected by 13 cfChIP-seq correlated with tissue contributions determined using bisulfite sequencing ( Figure 5C) 14 using data from a prior publication. The latter approach uses tissue-specific DNA methylation 15 signatures to assign tissue-specific cfDNA. Quantitatively, total cfDNA was 10 times higher in 16 transplant patients than healthy controls (Figure 5d). Cardiac-specific cfDNA was 6 times higher 17 in the heart transplant patients than healthy controls (Figure 5e), expected in these patients 18 where the heart allograft is exposed to host immunity directed against the allograft. To further 19 assess the greater heart injury in transplant patients, we measured allograft or donor-derived 20 cfDNA (ddcfDNA) by digital droplet PCR using primers directed to donor-recipient single 21 nucleotide polymorphisms. Again, we observed higher levels of ddcfDNA fraction (0.45%) in 22 . CC-BY 4.0 International license 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 made The copyright holder for this preprint this version posted January 25, 2023. ; https://doi.org/10.1101/2023.01.24.525414 doi: bioRxiv preprint transplant patients, compared to healthy controls who showed levels similar to background 1 (<0.01%). In addition, transplant patients showed higher tissue-specific cfDNA from 2 hematopoietic and other non-hematopoietic tissues ( (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 made The copyright holder for this preprint this version posted January 25, 2023. ; https://doi.org/10.1101/2023.01.24.525414 doi: bioRxiv preprint Discussion 1 In disease management, molecular of clinical and often molecular indicators is often needed to 2 guide treatment selection or monitor treatment response. Unfortunately, this remains a 3 limitation for many disease conditions such as transplantation where the existing tools are 4 invasive and/or have poor sensitivity (Marboe et al. 2005). This study shows that plasma cfDNA 5 chromatin immunoprecipitation sequencing (cfChIP-seq) may provide a minimally invasive 6 approach to identify biologically plausible gene and molecular pathway signals in both heathy 7 controls and heart transplant patients. The approach also profiles tissue-specific cfDNA and thus, 8 identifies tissue injury patterns that are biological plausible. We further demonstrate that cfChIP-9 seq approach provides replicable gene signals with different blood collection methods The study 10 findings are potentially broad applicability concurrent with a prior study. CC-BY 4.0 International license 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 made The copyright holder for this preprint this version posted January 25, 2023. ; https://doi.org/10.1101/2023.01.24.525414 doi: bioRxiv preprint post-transplantation when long-term complications and chronic rejection involving these 1 pathways develop. Additional studies are needed to define the relationship of these early gene 2 changes to chronic allograft fibrosis and chronic rejection, two long-term sequalae that involve 3 extracellular matrix remodeling. 4 The cfChIP-seq approach also identified tissue contributions of cfDNA that were consistent in 5 prior report (Sadeh et al. 2021). Heart transplant patients show higher cardiac specific-cfDNA 6 compared to healthy controls, expected in these patients. The higher cardiac injury correlated 7 with high allograft-derived cfDNA. We also observed elevated cfDNA from multiple 8 hematopoietic and non-hematopoietic tissue types. The tissue-injury pattern observed in heart 9 transplant patients is biological plausible and represent tissue types with increased incidence of

13
Studies with larger number of subjects will be relevant to validate this study findings and further 14 assess the clinical utility. These future studies should sequence cfDNA to higher sequencing depth 15 to fully assess the utility of cfChIP-seq signals for H3K4me1, H3K36me3 and other histone 16 antibodies. Studies assessing cfChIP-seq reproducibility following different sample storage 17 conditions are also needed. Pending these additional studies, this pilot study indicates that 18 cfChIP-seq reliable defines gene expression signals that are relevant in heart transplantation. 19 Such a minimally invasive approach may be utilized to reliable monitor transplant patients and 20 other patient population.

21
. CC-BY 4.0 International license 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 made The copyright holder for this preprint this version posted January 25, 2023. For each healthy control patient, two 8 -10 mL blood samples were collected simultaneously 13 into Streck cell-free DNA BCT tube (Streck) or into BD Vacutainer plastic EDTA tube to prepare 14 plasma. All peripheral blood samples were first centrifuged at 1,600 g for 10 min at 4 °C. Plasma 15 samples for half of the patients (both Streck and EDTA) underwent a second centrifugation at 16 3,000 g or 16,000 g for 10 min at 4 °C. One milliliter aliquots of plasma was stored at −80 °C until 17 use. For transplant patients, blood was collected into Streck tubes and centrifuged at 1,600 g 18 followed by a 16,000 g. One milliliter aliquots of plasma was stored at −80 °C until use. (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 made The copyright holder for this preprint this version posted January 25, 2023. ; https://doi.org/10.1101/2023.01.24.525414 doi: bioRxiv preprint cfDNA was estimated as the Alu247/Alu115 ratio and the recovery rate calculated with the 1 eluted/added ratio of λ-DNA by QPCR with specific λDNA primers (F'-2 CGGCGTCAAAAAGAACTTCC/R'-GCATCCTGAATGCAGCCATA). To prepare DNA library for input 3 cfDNA, Accel-NGS 2S plus DNA library kit was used with 1 ng of cfDNA with 9 cylces of PCR 4 amplification. The indexed DNA library was sequenced on Novaseq6000 as following standard 5 procedures (~1.4X10 8 properly aligned reads, 2X100 bp). (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 made The copyright holder for this preprint this version posted January 25, 2023. ; https://doi.org/10.1101/2023.01.24.525414 doi: bioRxiv preprint called using DiffBind v2 and its Deseq2 differential caller with default parameters. Peaks were 1 annotated using UROPA version 4.0.2 (Kondili et al. 2017) and Gencode Release 19 (GrCh37). 2 UROPA annotation conditions involved three query steps, each having an attribute.value filter of 3 "protein_coding" and a feature.anchor of "start". The three queries varied only by distance which 4 were set as: 3000, 10000, and 100000. For gene annotations, only the closest match (finalhit) 5 was used for downstream analyses. For promoter annotations, all genes that matched the first 6 query (allhit) were used for downstream analyses. Over-enrichment analysis was completed on 7 the promoter annotations for H3K4Me3 using clusterProfiler and plotted with enrichplot. 8 Specifically, the list of genes was compared to the KEGG human pathways with an appropriate 9 background set. 10 Statistical analysis and visualization 11 Graphs were generated by the R software (v3.6.3 or later) using Ggprism, Ggplot, Ggrepel, 12 patchwork, VennDiagram, EnrichedHeatmap, circlize, and karyoploteR. Python software (v3.5) 13 using pybedtools and pysam packages and GraphPad Prism software (v9.4.1) were also used. 14 Comparisons between groups were performed using nonparametric Mann-Whitney U test either 15 on GraphPad Prism or R software (v4.0.3) and adjusted for multiple testing using the Bonferroni 16 correction. P < 0.05 indicates statistically significant: * P < 0.05, * * P < 0.01, * * * P < 0.001, and 17 * * * * P < 0.0001. For the FRiP statistical test, differences FRiP between conditions were modeled 18 using linear mixed effects regression with no fixed effects (intercept-only) and a random effect 19 for bam files within each sample. A t-test was performed to assess statistical significance using 20 the nlme package in R.

21
. CC-BY 4.0 International license 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 made The copyright holder for this preprint this version posted January 25, 2023. Data and code availability 13 The authors declare that all data supporting the findings of this work are available within the 14 article and its supplementary information, and raw sequencing data are available from the 15 corresponding author upon reasonable request. All codes for cfChIP-seq data analysis is available 16 at github repository: https://github.com/OpenOmics/cfChIP-seek.

17
. CC-BY 4.0 International license 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 made The copyright holder for this preprint this version posted January 25, 2023. . CC-BY 4.0 International license 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 made The copyright holder for this preprint this version posted January 25, 2023. ; https://doi.org/10.1101/2023.01.24.525414 doi: bioRxiv preprint (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 made

Figure Legend
The copyright holder for this preprint this version posted January 25, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 signals between heart transplant and healthy controls, marked dots under the thick line depict 1 genes with lower H3K4Me3 signals in heart transplant subjects compared to healthy controls. (D) 2 KEGG pathway enrichment analysis of genes whose promoters were associated with a significant Calcineurin genes were defined as being members of GO:0097720 or Reactome R- HSA-2025928. 7 Genes above the thick line shows significant difference specific to transplantation state as 8 compared to healthy controls. (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 made The copyright holder for this preprint this version posted January 25, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 normalized reads/kb was multiplied with total cfDNA concentration and comparison was done 1 by Mann Whitney U test. P-value <0.05 was considered statistically significant; *p < 0.05; **p < 2 0.01; ***p < 0.001.

3
. CC-BY 4.0 International license 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 made The copyright holder for this preprint this version posted January 25, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023       Suppl. Figure 4. ChIP-seq tissue-specific signatures in different blood processing conditions. (A) 12 Comparison of tissue-specific signatures among the four processing conditions (E16, E3, S16, S3). 13 (B) The proportion of blood cells signatures between different conditions. 14 . CC-BY 4.0 International license 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 made The copyright holder for this preprint this version posted January 25, 2023. ; https://doi.org/10.1101/2023.01.24.525414 doi: bioRxiv preprint Figure S1 A B 0e+00 3e+07 6e+07 9e+07 0.0e+00 5.0e+07 1.0e+08 1.5e+08 Total reads    Table   1 Suppl. Table 1: Peaks, gene set with and without normalization for all four conditions 2 Suppl. Table 2: Housekeeping genes (also monocyte and neutrophil genes) 3 Suppl. Table 3: Number of differential genes by pathways 4 . CC-BY 4.0 International license 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 made