Colorectal adenocarcinoma metastasis to the liver samples from nine male patients aged 50–80 (mean, 65) years from the Gillies McIndoe Research Institute Tissue Bank were used for this study which was approved by the Central Health and Disabilities Ethics Committee (ref. no. 15/CEN/106). Written informed consent was obtained from patients included in this study.

Hematoxylin and eosin (H&E) staining was performed on 4μm-thick formalin-fixed paraffin-embedded sections of nine samples of CAML to confirm the presence of the tumor on the slides by an anatomical pathologist (HDB). 3,3-Diaminobenzidine (DAB) IHC staining was then performed on these sections for CD44 (1:1,500; cat# MRQ-13, Cell Marque, Rocklin, CA, USA), OCT4 (1:30; cat# MRQ-10, Cell Marque), SOX2 (1:200; cat# PA1-094, Thermo Fisher Scientific, Rockford, IL, USA), KLF4 (1:200; cat# NBP2-24749SS, Novus Biologicals LLC, Littleton, CO, USA), NANOG (1:100; cat# ab80892, Abcam, Cambridge, MA, USA), and c-Myc (1:1,000; ca# 9E10, Abcam) as previously described (22). All DAB IHC-stained slides were mounted in Surgipath Micromount (Leica, Nussloch, Germany). To confirm coexpression of two proteins, two representative samples of CRCML from the original cohort of nine samples used for DAB IHC staining underwent immunofluorescence (IF) IHC staining. Vectafluor Excel antimouse 488 (ready-to-use; cat#VEDK2488, Vector Laboratories, Burlingame, CA, USA) and Alexa Fluor antirabbit 594 (1:500; cat#A21207, Life Technologies, Carlsbad, CA, USA) were used to detect the combinations. All IF IHC-stained slides were mounted in Vecta Shield Hardset mounting medium with 4′,6′-diamino-2-phenylindone (Vector Laboratories). All antibodies were diluted in Bond primary diluent (Leica). All DAB and IF IHC staining was performed using the Leica Bond Rx auto-stainer (Leica), as previously described (22).

Positive human control tissues used for the primary antibodies were seminoma for OCT4 and NANOG (23), skin for SOX2 (24), breast cancer for KLF4 (25) and prostate for c-Myc (26). A negative CAML control sample was prepared for DAB IHC staining using an IgG isotype control (ready-to-use; cat#IR600, Dako, Santa Clara, CA, USA). Negative controls for IF IHC staining was performed using a section of glioblastoma tissue with the combined use of primary isotype mouse (ready-to-use; cat# IR750, Dako, Copenhagen, Denmark) and rabbit (read-to-use; cat# IR600, Dako) antibodies.

3,3-Diaminobenzidine IHC-stained slides were viewed and images were captured using the Olympus BX53 microscope fitted with an Olympus DP21 digital camera (Olympus, Tokyo, Japan). IF IHC-stained slides were viewed and imaged using the Olympus FV1200 biological confocal laser-scanning microscope and processed with cellSens Dimension 1.11 software using 2D deconvolution algorithm (Olympus).

RNA was extracted from six snap-frozen samples of CAML from the same cohort of nine patients used for DAB IHC staining, was analyzed using NanoString nCounter™ Gene Expression Assay (NanoString Technologies, Seattle, WA, USA). Total RNA was extracted using the MagJET RNA kit (Thermo Fisher Scientific) with the protocol adapted for tissue, and run on a KingFisher Duo machine (Thermo Fisher Scientific). The RNA samples were then quantitated on a Qubit^® 2.0 fluorometer (Invitrogen, Life Technologies) and were subject to RNA integrity analysis via the 2100 Bioanalyzer Instrument (Agilent Technologies). The samples then underwent NanoString nCounter gene expression assay performed by New Zealand Genomics (Dunedin, New Zealand) according to the manufacturer’s protocol. Probes for the genes encoding OCT4 (POU5F1, NM_002701.4), SOX2 (NM_003106.2), NANOG (NM_024865.2), KLF4 (NM_004235.4), c-Myc (NM_002467.3), and the housekeeping gene GUSB (NM_000181.1) were designed and synthesized by NanoString Technologies. Raw data were analyzed using nSolver™ software (NanoString Technologies) using standard settings and normalized against the housekeeping gene.

The 4μm-thick formalin-fixed paraffin-embedded CAML sections of six patients from the original cohort of nine patients used for DAB IHC staining were used for CISH staining using the Leica Bond Rx autostainer and detected using the ViewRNA eZ Detection kit (Affymetrix, CA, USA) as previously described (27). Probes for the stem cell markers c-Myc (NM_002467), KLF4 (NM_004235), NANOG (NM_024865), SOX2 (NM_003106), and OCT4 (POU5F1, NM_002701) were obtained from Affymetrix. Positive controls were demonstrated on human tissues: breast cancer for KLF4, normal prostate for c-Myc, skin for SOX2, and seminoma for OCT4 and NANOG. Negative control was demonstrated on a sample of Bacillus (NM_L38424).

Counting of cells with nucleus or cytoplasm that stained positively for OCT4, SOX2, NANOG, KLF4, and c-Myc was each performed in six fields of view at 400× magnification representative of both tumor nests (TNs) and the peritumoral stroma for each of the original nine samples of CAML used for DAB IHC staining. In each field of view, cells within the TNs and those within the peritumoral stroma were counted separately and hence images were taken to incorporate areas of tissue containing cells both within the TNs and the peritumoral stroma.

Cell counting for CISH-stained slides was performed at 1,000× magnification which meant that cells within the TNs and those within the peritumoral stroma could not be incorporated in a single image and hence images were taken separately for cells within the TNs and those within the peritumoral stroma.

Statistical analysis for comparison of markers demonstrated by DAB IHC and CISH staining was performed to determine significance. Total counts of cells that stained positively and those that stained negatively for each ESC marker within the TNs and the peritumoral stroma were collated. The proportion of positive staining cells was calculated for each marker in both the TNs and peritumoral stroma. χ² tests were then carried out using the Quantpsy (www.quantpsy.org) software to determine statistical significance with regards to the proportional expression of each marker, and related t-tests were used to compare the proportional expression of each ESC marker within the TNs against those within the peritumoral stroma using SPSS V.24.

Hematoxylin and eosin staining confirmed the presence of CAML on the slides, distinguishing CAML from normal liver architecture and identifying two distinct populations of cells within adenocarcinoma glandular TNs and the peritumoral stroma (Figure S1A in Supplementary Material).

Membranous staining of CD44 was demonstrated on the cells within the TNs (Figure S1B in Supplementary Material, brown). Cytoplasmic staining for OCT4 (Figure S1C in Supplementary Material, brown) was observed on cells within the peritumoral stroma but not those within the TNs, with the cells that stained positively appearing in distinct clusters throughout the tissue. NANOG (Figure S1D in Supplementary Material, purple), SOX2 (Figure S1E in Supplementary Material, brown), and c-Myc (Figure S1F in Supplementary Material, brown) were expressed on the cells within the TNs (Figures S1D–F in Supplementary Material, arrows) and peritumoral stroma (Figures S1D–F in Supplementary Material, arrowheads). Cytoplasmic staining of KLF4 (Figure S1G in Supplementary Material, purple, arrows) was present on cells within the TNs (Figure S1G in Supplementary Material, arrows), and to a lesser extent, in cells within the peritumoral stroma (Figure S1G in Supplementary Material, brown, arrowheads).

Normal positive staining patterns for OCT4 (Figure S2A in Supplementary Material, brown), NANOG (Figure S2B in Supplementary Material, purple), SOX2 (Figure S2C in Supplementary Material, brown), c-Myc (Figure S2D in Supplementary Material, brown), and KLF4 (Figure S2E in Supplementary Material, purple) were demonstrated on human seminoma, normal skin, breast cancer, and prostate tissues, respectively. The negative control was a CAML sample DAB IHC stained using an IgG isotype (Figure S2F in Supplementary Material).

To demonstrate coexpression of the ESC markers, IF IHC staining was performed on two representative CAML samples from the original cohort of nine patients used for DAB IHC staining. SOX2 (Figures 1A,D, red), NANOG (Figures 1B,E, red), KLF4 (Figures 1C,F, red), and c-Myc (Figures 1D–F, green) were expressed on cells within the TNs (Figures 1A–F, thin arrows) and the peritumoral stroma (Figures 1A–F, arrowheads). OCT4 (Figures 1A–C, green) was expressed on the SOX2⁺ (Figure 1A, green, arrowheads), NANOG⁺ (Figure 1B, green, arrowheads), and KLF4⁺ (Figure 1C, green, arrowheads) cells within the peritumoral stroma. Furthermore, there were cells within the peritumoral stroma that expressed SOX2 (Figure 1A, red, thick arrows), NANOG (Figure 1B, red, thick arrows), and KLF4 (Figure 1C, red, thick arrows) that did not express OCT4.

To determine the expression of c-Myc (Figures 1D–F, green), we performed dual staining with SOX2 (Figure 1D, red), NANOG (Figure 1E, red), and KLF4 (Figure 1F, red). c-Myc was expressed by cells within the TNs (Figures 1D–F, green, arrows) and cells within the peritumoral stroma (Figures 1D–F, arrowheads). Intriguingly the c-Myc⁺ population within the peritumoral stroma also expressed SOX2 (Figure 1D, red), NANOG (Figure 1E, red), and KLF4 (Figure 1F, red).

Split images of all the stains presented in Figure 1 are shown in Figure S3 in Supplementary Material. Minimal staining was present on the negative control (Figure S3M in Supplementary Material), confirming the specificity of the primary antibodies used.

NanoString mRNA analysis confirmed expression of mRNA transcripts for OCT4, NANOG, KLF4, and c-Myc in all six CAML samples relative to the housekeeping gene GUSB (Figure 2). SOX2 mRNA transcript expression was below detectable levels in all CAML samples. RNA integrity analysis confirmed that two of the four samples had RIN > 7.

Colorimetric in situ hybridization demonstrated the presence of mRNA transcripts for OCT4 (Figure S4A in Supplementary Material, pink, arrows), SOX2 (Figure S4B in Supplementary Material, pink, arrows), NANOG (Figure S4C in Supplementary Material, pink, arrows), KLF (Figure S4D in Supplementary Material, pink, arrows), and c-Myc (Figure S4E in Supplementary Material, pink, arrows) in all six samples of CAML. Positive controls were human seminoma for OCT4 (Figure S5A in Supplementary Material, pink, arrows), normal skin for SOX2 (Figure S5B in Supplementary Material, pink, arrows), seminoma for NANOG (Figure S5C in Supplementary Material, pink, arrows), breast cancer for KLF4 (Figure S5D in Supplementary Material, pink, arrows), and normal prostate for c-Myc (Figure S5E in Supplementary Material, pink, arrows). The negative control was a sample of Bacillus (NM_L38424) (Figure S5F in Supplementary Material).

Statistical analyses of the results of cell counting of DAB-stained slides showed that there was statistically significant difference between the mean proportions of OCT4⁺, SOX2⁺, NANOG⁺, and KLF4⁺ cells in the peritumoral stroma compared to those within the TNs (p < 0.05), but there were no significant differences for c-Myc (p > 0.05) (Figure 3A). When comparing the total proportional of positively stained cells within the TNs and the peritumoral stroma for each marker, χ² statistical analysis demonstrated a hierarchy of expression of these markers with increasing abundance: NANOG > SOX2 > KLF4 > c-Myc > OCT4. All differences were statistically significant (p < 0.01).

The means for the cell counts on CISH-stained slides (Figure 3B) analyzed by χ² statistical analysis demonstrated the following hierarchy: c-Myc > KLF4 > NANOG > OCT4 > SOX2. All comparisons were highly statistically significant between markers (p < 0.0001) except for the comparison between KLF4 and NANOG which was not statistically significant (p = 0.884).