Ethical statement
Ethical clearance for the collection of stored normal colorectal tissue biopsies from healthy individuals (serving as control tissues), as well as tumor tissues from patients with newly diagnosed colorectal adenocarcinoma was obtained from the Health Research Ethics Committee (HREC) of Stellenbosch University (ethics reference: 6585). All study participants signed an informed consent form prior to participation in the study and sample collection. This study, including sample collection and sample processing, was conducted according to the guidelines set by the Declaration of Helsinki.
Sample collection and study population
Table 1 shows the sample demographics of healthy and CRC populations. Formalin-fixed, paraffin-embedded (FFPE) morphologically normal colon biopsies from 10 healthy individuals and FFPE colorectal tumor resections (n=18) and biopsies (n=6) from 24 patients with newly diagnosed (histologically confirmed) colorectal adenocarcinoma were collected from the National Health Laboratory Service at Tygerberg Hospital, Cape Town, South Africa. The stage of all recruited patients was assessed, with stage 1 (n=1), stage 2 (n=10), stage 3 (n=8), and stage 4 (n=5) CRC patients included. The American Joint Committee on Cancer/Union for International Cancer Control (AJCC/UICC) 8th edition staging system was used to assess the stage of patients who had tumor resections, while CT scans were used to assess the stage of patients who had biopsies taken. None of the CRC patients had any cancer treatment at the time of sample collection (no neoadjuvant chemotherapy or radiotherapy). Genetic predisposition did not form part of the exclusion criteria.
Table 1. Demographic information of healthy individuals and colorectal cancer (CRC) patients.
Demographics
|
|
Healthy individuals
|
CRC patients
|
Gender
|
Male (n=5), Female (n=5)
|
Male (n=11), Female (n=13)
|
Age (years)
|
60.5 [37.75-69.25]
|
55.5 [44-68.5]
|
Data expressed as median and [25% - 75% quartile range].
Tissue preparation
An automated tissue processer (Tissue-Tek® VIP™ Vacuum Infiltration Processor) was used for tissue processing, which runs for a total time of ± 12 hours according to the protocol in Supplementary Table S1. Following sectioning of blocks (using the Leica RM 2125 rotary microtome (SMM instruments, Germany)), tissue sections were transferred to a water bath and placed onto standard microscope slides for haematoxylin and eosin (H&E) staining, or positively charged microscope slides for immunohistochemistry and for confocal analysis. Prior to staining, slides were incubated at approximately 60˚C for about one hour for the removal of wax. Ultimately, following staining of the sections, a glass coverslip was mounted onto the tissue using distyrene, a plasticizer, and xylene (DPX) mounting media and left to dry for 48 hours (Dako fluorescence mounting medium was used for confocal analysis).
Haematoxylin and eosin (H&E) staining
To study the morphology and structure of healthy colorectal tissues and CRC tumor tissues, the H&E stain was used to stain tissue sections. Slides were placed into a plastic staining rack, which was placed into the Leica Auto Stainer XL (SMM instruments, Germany). The autostainer follows a pre-programmed procedure, including steps for the deparaffinisation, rehydration, and clearing of the tissue (see Supplementary Table S2). Zeiss Axioskop 2 Microscope (Carl Zeiss, Germany) was used to view the slides and examine staining, with Zen Lite software (v2.3, Germany) used to capture the images.
Immunohistochemistry
All immunohistochemical staining procedures were performed on the BOND-MAX automated system (Leica, Wetzlar, Germany), using the Bond Polymer Refine Detection System (Leica BondTM) (Cat no. DS9800). An automated immunohistochemical staining protocol was used, including a standard dewax (using Bond Dewax Solution (Leica BondTM) (Cat no. AR9222)) and rehydrate program, as well as pre-treatment for antigen retrieval. Refer to Table 2 for the automated staining protocol. Tissues were either incubated in Bond Epitope Retrieval (ER) Solution 2 (Leica BondTM) (Cat no. AR9640) or Bond ER Solution 1 (Leica BondTM) (Cat no. AR9961), prior to incubation with the primary antibody. Following a wash step (standard wash protocol with Bond Wash Solution (Leica BondTM) (Cat no. AR9590)), endogenous peroxidases in tissue sections were blocked. After incubation with the primary antibody, the chromogen diaminobenzidine (DAB) was used to visualize positive antibody-antigen reactions, and counterstained with haematoxylin to observe tissue morphology. Following completion of automated staining, the samples were dehydrated (refer to Supplementary Table S3). Zeiss Axio Observer 7 inverted Microscope (Carl Zeiss, Germany) with Axiocam 305 colour camera, using the brightfield modality and 5x objective, was used to view the slides and examine staining.
Immunohistochemical protocols were developed and optimized to detect the proteins of interest in the tissue samples. The following antibodies were used: anti-H. pylori antibody, anti-E. coli antibody, anti-E. coli LPS antibody, and anti-SAA antibody. The optimal dilution factor of each antibody (step 9 in Table 2) (all antibodies were diluted in Bond Primary Antibody Diluent (Leica BondTM) (Cat no. AR9352)) and the optimal ER buffer in each case (step 3 in Table 2) were optimized. Positive (tissue) controls for H. pylori, E. coli, LPS, and SAA were used to confirm that each antibody binds to its specific immunogen. Negative antibody controls were also included. Negative and positive control images are shown in Supplementary Fig. S1-S4. Zeiss Axioskop 2 Microscope (Carl Zeiss, Germany) was used to view these control slides and examine staining, with Zen Lite software (v2.3, Germany) used to capture the images.
To detect H. pylori in tissue sections, Anti-Helicobacter pylori antibody [EPR10353] (rabbit monoclonal IgG, ab172611, Abcam) was used (diluted 1:250). ER1 buffer was used for heat-induced epitope retrieval (HIER). H. pylori human gastritis tissue (from a stomach biopsy), as recommended by the antibody datasheet, was used as a positive tissue control. For detection of E. coli, E. coli serotype 0157 Monoclonal Antibody (I88H) (mouse monoclonal IgG, MA1-7303, Invitrogen) was used (diluted 1:100), using ER2 buffer. For a positive control, healthy platelet poor plasma (PPP) was exposed to a colony of E. coli cells (E. coli ATCC 13706). The Cytospin® 3 Cell Preparation System was used to deposit a monolayer of cells in a defined area onto a microscope slide, using centrifugal force. A volume of 150 μL of the exposed PPP sample was added to the sample chamber, and centrifuged at a speed of 2000 rpm for 3 minutes. To detect bacterial LPS in tissue sections, Anti-E. coli LPS antibody [2D7/1] (mouse monoclonal IgG, ab35654, Abcam) was used (diluted 1:100), using ER2 buffer. Rat intestinal tissue, as recommended by the antibody datasheet, was used as a positive tissue control. To detect SAA in tissues, immuno-purified rabbit anti-human SAA antibody (kindly provided by Prof. Frederick C. de Beer and Dr. Marcielle de Beer from the University of Kentucky), with a concentration of 4.1 mg mL-1, was used (diluted 1:1500), using ER1 buffer. Human tonsil tissue was used as a positive tissue control.
Table 2. Automated immunohistochemical staining protocol.
Step
|
Reagent
|
Time
|
Temperature
|
1
|
Bake
|
Standard bake
|
30 minutes
|
60°C
|
2
|
Dewax, rehydrate
|
Bond Dewax Solution, 100% ethanol and deionized water
|
|
|
3
|
Antigen retrieval
|
ER1 or ER2 buffer
|
20 minutes
|
100°C
|
4
|
Wash
|
Bond Wash Solution
|
0 minutes
|
|
5
|
Block endogenous peroxidase
|
Peroxide block (from Bond Polymer Refine kit)
|
5 minutes
|
23°C
|
6
|
Wash
|
Bond Wash Solution
|
0 minutes
|
|
7
|
Bond Wash Solution
|
0 minutes
|
|
8
|
Bond Wash Solution
|
0 minutes
|
|
9
|
Primary antibody binding
|
Primary antibody diluted in Bond Primary Antibody Diluent
|
15 minutes
|
23°C
|
10
|
Wash
|
Bond Wash Solution
|
0 minutes
|
|
11
|
Bond Wash Solution
|
0 minutes
|
|
12
|
Bond Wash Solution
|
0 minutes
|
|
13
|
Detection of bound antibody
|
Post Primary (from Bond Polymer Refine kit)
|
8 minutes
|
23°C
|
14
|
Bond Wash Solution
|
2 minutes
|
|
15
|
Bond Wash Solution
|
2 minutes
|
|
16
|
Bond Wash Solution
|
2 minutes
|
|
17
|
Polymer (from Bond Polymer Refine kit)
|
8 minutes
|
23°C
|
18
|
Wash
|
Bond Wash Solution
|
2 minutes
|
|
19
|
Bond Wash Solution
|
2 minutes
|
|
20
|
Deionized water
|
0 minutes
|
|
21
|
Deionized water
|
0 minutes
|
|
22
|
Colour development
|
DAB (from Bond Polymer Refine kit)
|
10 minutes
|
23°C
|
23
|
Wash
|
Deionized water
|
0 minutes
|
|
24
|
Deionized water
|
0 minutes
|
|
25
|
Deionized water
|
0 minutes
|
|
26
|
Counterstain
|
Haematoxylin (from Bond Polymer Refine kit)
|
5 minutes
|
|
27
|
Wash
|
Deionized water
|
0 minutes
|
|
28
|
Deionized water
|
0 minutes
|
|
29
|
Deionized water
|
0 minutes
|
|
To simplify automated imaging of the tissue randomly positioned on the different slides, a block was drawn around each tissue section to fully encompass the outer boundaries of the specific section. The coordinates of the four corners of this box shaped perimeter were then used to set up a tile scan in the Zen Pro software (v2.5, Germany). The software automatically calculated the number of images to acquire to scan the complete section. Importantly, tiles containing artefacts were excluded from analysis, as well as tiles containing no tissue in the field of view. About 97 tile images were analysed per CRC tumor tissue sample, and ± 11 tile images per healthy colorectal tissue sample. The stitching function of the analyses section of the Zen software, which allows for the correct alignment of the tiles to view a complete section, was applied. However, only the individual tiles were used for statistical analyses. The intensity and exposure time settings were 8.3% and 0.08 seconds, respectively.
Confocal analysis for the detection of structural protein changes
In order to investigate structural protein changes present in healthy colorectal tissues and CRC tumor tissues, tissue sections were incubated with the amyloid-selective AmytrackerTM 630 marker (Ebba Biotech AB). Tissue sections were either stained with only the fluorescent amyloidogenic marker, or with both AmytrackerTM 630 and Hoechst fluorescent dye (supplied at 10 mg mL-1 (Sigma)). First, tissue sections were deparaffinised (refer to Supplementary Table S4) and fixed with ice-cold ethanol (95%) for 5 minutes at room temperature. Following rehydration of the tissue sections in a mix of ethanol and deionized water (1:1) for 5 minutes, sections were equilibrated in Gibco™ phosphate-buffered saline (PBS) (pH=7.4) (ThermoFisher Scientific, 11594516) for 5 minutes. For staining with only the AmytrackerTM 630 marker, tissue sections were subsequently incubated in AmytrackerTM 630 (diluted 1:1000 in PBS) for 30 minutes, followed by PBS wash steps (2x5 minutes). In cases where tissue sections were also stained with Hoechst fluorescent dye, following this wash step, tissue sections were incubated in Hoechst dye (diluted 1:200 in PBS) for 5 minutes and then further washed in PBS (3x5 minutes). Single stained and unstained samples were included for method development and optimization of staining of tissue sections with both AmytrackerTM 630 and Hoechst dye. A Zeiss LSM 780 with ELYRA PS1 confocal microscope with a Plan-Apochromat 20x/0.8 objective was used to view the slides.
The following settings were used:
- Staining with only AmytrackerTM 630: 488 nm excitation laser, with emission measured at 606–695 nm
- Staining with AmytrackerTM 630 and Hoechst fluorescent dye:
- Amytracker™ 630: 488 nm excitation laser, with emission measured at 606–695 nm
- Hoechst fluorescent dye: 405 nm excitation laser, with emission measured at 415–460 nm
For data acquisition and comparison between healthy colorectal tissue and CRC tumor tissue images, gain settings were kept constant to ensure a reliable statistical outcome. All micrographs were captured in an unbiased manner as 3x3 tile images. At least 4 tile images were captured per sample stained with Amytracker™ 630. Of 4 healthy colorectal tissue samples, 17 images were acquired, with 23 images acquired from 4 CRC tumor tissue samples. The mean fluorescence intensity (MFI) of all acquired images was computed (using ImageJ (FIJI) 34) to compare the response to the Amytracker™ 630 marker in control tissues and CRC tumor tissues. In addition, unstained healthy colorectal tissue samples and unstained CRC tumor tissue samples were analysed to account for the autofluorescence of the tissues. A minimum of 2 tile images were captured per unstained sample. The average MFI of the corresponding unstained CRC tumor tissue samples was subsequently subtracted from the individual MFI values of the CRC tumor tissue samples stained with Amytracker™ 630. The same approach applied to the healthy colorectal tissue samples.
Statistical analysis
The immunohistochemistry signal was analysed in the following straightforward manner. 1D colour histograms for each of the four antibody signals were formed based on manually identified regions across multiple images. These histograms are shown in Figure 7A. These (normalized) histograms were then applied across each image to determine a total signal strength. A logistic statistical model (using glm in R 4.0.2) was then applied to determine the strength of association (and confidence intervals) between the signal strength and disease status. As the dynamic range of the signal is very large, a log transform was performed to the signal strength. Odds Ratios are reported on the log scale after z-score standardisation to allow meaningful comparison between stains. Lastly, as the datasets were imbalanced, with more disease images than control images, we performed an additional analysis with inverse frequency weighting of the logistic loss. Results are consistent between these weighted and unweighted models with the exception of H. pylori, where we make more conservative claims. The unweighted model does not account for the imbalance regarding the number of control tissue and CRC tumor tissue samples, and also not for the imbalance in the number of images per sample, while the weighted model accounts for these data imbalances (making it unbiased).
Data of the confocal analysis were tested for normality using the Shapiro-Wilk normality test, and analysed using the Mann-Whitney non-parametric test in GraphPad Prism 7.04 (the image plot of the confocal data was prepared in GraphPad Prism 7.04). Statistical significance was accepted at p<0.05.