Discovering CsgD Regulatory Targets in Salmonella Biofilm Using Chromatin Immunoprecipitation and High-Throughput Sequencing (ChIP-seq)

Preparation of chromatin immunoprecipitation samples from bacterial biofilm involves several steps, as shown in Figure 1. These include growing biofilm in flask culture, collecting conditioned media, collecting an adequate amount of biofilm and planktonic cells as control, washing cells in PBS, homogenization, crosslinking proteins to DNA, lysing cells, fragmenting DNA by sonication, immunoprecipitating DNA with the appropriate antibodies and protein G beads, washing and eluting immunoprecipitated DNA, and purifying DNA for library preparation and sequencing.

S. Typhimurium cells grown in liquid culture under biofilm-inducing conditions undergo phenotype switching to form multicellular biofilm aggregates and planktonic cells. This population will contain approximately 40% biofilm aggregates, which express higher levels of diguanylate cyclases (DGCs) and biofilm regulators, and 60% planktonic cells, which express higher levels of virulence-associated genes^2,4 (Figure 2A). In this protocol, we describe a method to harvest both cell types to compare transcription sites through ChIP-seq; however, this protocol can be adapted for other types of biofilms formed by other bacterial species. Biofilm aggregates and planktonic cells are separated by slow speed centrifugation, which pellets biofilm and leaves planktonic cells in the supernatant² (Figure 2B). The cell types are then treated separately for subsequent steps in the protocol.

The recommended amount of DNA input for ChIP-seq is 10-25 µg²⁰. In S. Typhimurium flask culture, 30 mg of biofilm yields approximately 25 µg DNA. Since biofilm aggregates have an abundance of proteinaceous extracellular material, we hypothesized that this would interfere with the efficiency of cross-linking. If we were simply to normalize the different cell types by cell number, treatment of planktonic cells may result in an unequal amount of crosslinked product presented for immunoprecipitation. A protein assay was performed to determine the equivalent amount of planktonic cell material to match 30 mg of biofilm (Figure 3). 6.0 OD₆₀₀ of planktonic cells were harvested to match 30 mg biofilm.

Biofilm aggregates must first be broken apart to allow equal access of cross-linker to all cells. We found that the most uniform and high-throughput way to homogenize the cells was using a mixer mill and 5 mm stainless steel bead (Figure 4A). Planktonic cells are treated the same way to reduce variables in the ChIP-seq experiment (Figure 4B).

Once the DNA has been fragmented by sonication, crosslinked, and treated with RNase and Proteinase K, it can be visualized on a 2% agarose gel. Sonicated DNA is broken down into a collection of fragments that appear as a smear on the agarose gel. The average fragment size decreases with an increasing number of sonication bursts (Figure 5). Energy transfer will vary for different sonicator units, so a sonication assay is recommended to determine the number of sonication bursts required to fragment DNA to an average of less than 1 kb. For our ChIP-seq experiment, we chose 5 bursts.

ENCODE guidelines recommend a confirmation of antibody target-binding activity with an immunoblot²¹. In this case we measured our antibody's specificity and strength of binding by immunoblot on whole cell lysates prepared for ChIP (Figure 6). We confirmed that the antibody binds to CsgD-3xFLAG; anti-FLAG and anti-CsgD signals colocalize at the expected molecular weight (28 kDa)²².

ChIP procedures often yield low concentrations of DNA. Therefore, a purification method that removes contaminants and retains a high proportion of the DNA was chosen. Magnetic bead purification was chosen over column-based purification because it retained low concentrations of input ChIP DNA better (Figure 7) and removed contaminants (data not shown).

Due to reduced starting amounts of DNA, library preparation of ChIP DNA may require diluted adapters and PCR enrichment. Prior to sequencing, libraries can be visualized by Bioanalyzer trace to confirm correct size distribution before and after PCR enrichment (Figure 8). Additionally, if there is a known regulatory target of the transcription factor, qPCR should be used to confirm enrichment of binding regions by comparing fold change (∆∆Ct) between known target and reference gene sequence abundance in immunoprecipitated and sonicated input DNA.

ChIP-sequencing data should be analyzed by assigning quality scores, trimming low quality bases, aligning forward and reverse reads (in paired end sequencing), assembling to a high-quality reference genome, finding peaks above background, and performing downstream analysis (Figure 9A). Downstream analysis should include visualization and annotation of peaks, consistency with biological replicate samples, and motif analysis, and could include differential binding analysis between conditions or cell types, and data integration with gene expression from known pathways. For ChIP-seq data such as ours, peaks should be identified as significantly above background (Figure 9B,C, red bars) with a p value determination, not simply by fold-change. In the final analysis, the mapped reads are visualized against the annotation of the reference genome (Figure 9D). A poor ChIP-seq result would appear like input-seq without any significant peaks above background.

Figure 1: Illustration of steps in ChIP-seq with bacterial biofilms. In the described procedure, S. Typhimurium cells are grown in 1% Tryptone flask culture at 28 °C with shaking (1), cell-free conditioned media is collected for handling biofilm cell types (2), which is used to collect an adequate amount of biofilm and planktonic cells (3). These cells are washed with PBS and homogenized to break apart biofilm (4). Proteins are crosslinked to DNA with a formaldehyde crosslinker, after which cells are lysed to release cell contents (5). DNA in the cell lysate is fragmented by sonication (6). DNA crosslinked to the target protein is selected through immunoprecipitation with an antibody that binds to the target protein and Protein G magnetic beads (7). Selected DNA is washed and eluted from Protein G magnetic beads (8) and purified for library preparation and sequencing (9). Please click here to view a larger version of this figure.

Figure 2: In vitro flask model for studying S. Typhimurium biofilm development. (A) S. Typhimurium cells grown in 1% tryptone for 13 h at 28 °C undergo phenotype switching to form biofilm aggregates and planktonic cells in the liquid phase of the flask culture. (B) Biofilm aggregates and planktonic cells from the flask culture in (A) were separated through centrifugation at 210 x g for 2 min. The supernatant was harvested for planktonic cell preparations and the pellet was harvested for biofilm cell preparations. Please click here to view a larger version of this figure.

Figure 3: Total protein concentrations for planktonic cells and biofilm cell samples harvested from S. Typhimurium flask culture. Colorimetric protein assays were performed and protein amounts relative to a BSA standard curve were measured at 750 nm. Protein content of lysed planktonic cell samples at 4.0, 6.0, and 8.0 OD₆₀₀ were compared to 30 mg of biofilm aggregates. Please click here to view a larger version of this figure.

Figure 4: Homogenization of cell samples from S. Typhimurium biofilm flask cultures. (A) Biofilm aggregates from flask culture resuspended in PBS (left) were homogenized using a mixer mill at 30 Hz for 5 min (right). (B) Planktonic cells from flask culture resuspended in PBS were processed by mixer mill at 30 Hz for 5 min to ensure consistency between cell type samples. Please click here to view a larger version of this figure.

Figure 5: Sonication assay with pre-ChIP cell lysates. Biofilm aggregate and planktonic cell samples were separated, homogenized, crosslinked, and lysed. DNA-protein complexes were sonicated up to 8 times at 30 s on and 2 min off on ice to break DNA into smaller fragments. A portion of the sonicated lysate was separated on a 2% agarose gel. Please click here to view a larger version of this figure.

Figure 6: Target-binding specificity of anti-FLAG antibody used in ChIP-seq. Antibody specificity was assessed by immunoblotting against whole cell lysates prepared from flask cultures of the S. Typhimurium strains as shown. Strain ∆csgD + p3xFLAG/csgD and WT biofilm samples represent biofilm aggregates harvested from flasks, whereas strain ∆csgD ± p3xFLAG and WT planktonic cells represent planktonic cells. Purified His-tagged CsgD was loaded as a control. Whole cell lysates were normalized so that 30 µg of total protein was analyzed in each lane. Numbers on the left refer to the size (in kDa) of the prestained molecular weight markers. Primary antibody used for the upper blot was rabbit-anti-FLAG polyclonal antibody (Sigma-Aldrich #F7425), whereas the lower blot was incubated with rabbit-anti-FLAG together with a CsgD-specific monoclonal antibody². Goat anti-rabbit IgG (LiCor IRDye 680RD, 925-68071) and goat anti-mouse IgG (LiCor IRDye 800CW, 925-32210) were used as secondary antibodies. Fluorescent signals were visualized using the Odyssey CLx imaging system and Image Studio 4.0 software package (Li-Cor Biosciences). Representative images are shown. Please click here to view a larger version of this figure.

Figure 7: Column- or magnetic bead-based purification strategies for small amounts of DNA resulting from ChIP-seq experiments. Samples of known quantities of DNA were purified using column-based kits or magnetic beads and the concentration of the eluate was measured post-purification. Magnetic beads showed better recovery at low DNA levels, similar to the low amounts of DNA typically encountered at the end of ChIP-seq protocol, but prior to NGS library preparation. Please click here to view a larger version of this figure.

Figure 8: ChIP DNA preparations and subsequent libraries visualized by Bioanalyzer. (A) Average fragment size of genomic DNA after cell lysis and sonication was <1000 bp, with a majority of fragments in the 200–500 bp range. (B) Starting material for library preparation was below detection. (C) Adapter ligation and PCR enrichment allows for amplification of library fragments. (D) A poor library preparation has low DNA peaks, odd shaped or high molecular weight peaks, or abundant adapter peaks at approximately 100 bp. Please click here to view a larger version of this figure.

Figure 9: ChIP-seq data is analyzed using bioinformatic tools and visualized on a genome browser. A. Flowchart for typical bioinformatic analysis of ChIP-seq data. Raw reads for biofilm (∆csgD strain + p3xFLAG/csgD) and planktonic cells (∆csgD + p3xFLAG) were cleaned for low-quality reads and adapters, and mapped to S. Typhimurium 14028s genomes (NC_016856.1 for the chromosome and NC_016855.1 for the plasmid) by Bowtie2 (v2.3.3.1) with default parameters²³. MACS2 (v2.1.2) was used to call the peaks with parameters of '-q 0.01 — nomodel', taking biofilm replicates as testing datasets and planktonic ones as control²⁴. The MACS2 bdgcmp module was further used to generate fold-enrichment track with parameters of '-m FE'. The significant peak file with fold-enrichment track was transformed to a wig file using bedtools/bedClip/bedGraphToBigWig²⁵ and visualized on the reference genome using Integrative Genomics Viewer v2.5.1²⁶. Representative peaks are shown (red bars). (B) Large region of ~40 kbp, which contains multiple peaks, is an atypical result but still potentially valuable. (C) This is a more typical result, showing two significant peak regions. (D) Zoom in on left peak in C, showing reads mapping to the region of the S. Typhimurium 14028s genome containing divergent promoters for uspA and uspB. Please click here to view a larger version of this figure.

Table 1. Buffer recipes.