A total of 51 patients with CRC who underwent resection of the primary tumor were enrolled between January 2014 and December 2016 at the Autonomous University of Nuevo Leon, University Hospital ‘Dr. Jose Eleuterio Gonzalez’ (Monterrey, Mexico) and the High Specialty Medical Unit (UMAE 25) of the Mexican Social Security Institute (IMSS) and the Century XXI National Medical Center, IMSS. This project was authorized by the Ethics Committee of the Autonomous University of Nuevo Leon Medical School and University Hospital (approval no. BI14-009) and by the National Committee of Bioethics of the IMSS (approval no. R-2012-785-075). All participants signed an informed consent letter.

The inclusion criteria included the following: Men and women aged >18 years with a diagnosis of CRC who wished to participate in the study and who signed the informed consent letter. As part of the study participants had to undergo CRC surgical removal. The exclusion criteria included the following: Patients with a family history of CRC who did not sign the informed consent letter, pregnant women, Karnofsky Scale <60. Clinical characteristics and sociodemographic data are shown in Table I. The quality of life of the participants was valued immediately before surgery and 24 months after starting treatment using the Karnofsky Scale (27).

Fresh sample fragments (>1.4 cm²) were obtained from CRAC, normal tissue adjacent to the tumor (NAT; 3–7 cm distance from tumor) and normal colon from a donation of the historical collection of unidentifiable samples from the Department of Forensic Medicine of the University Hospital of the Autonomous University of Nuevo Leon. Normal colon tissues were preserved in RNA later stabilization solution at −20°C (Thermo Fischer Scientific, Inc.). Each sample was then divided into four fragments. A random fragment was used for chemosensitivity assays in a primary culture assay. A second fragment was fixed with Carnoy's solution, which contained ethanol, chloroform and glacial acetic acid (6:3:1 ratio), for 24 h at room temperature, embedded in paraffin, cut into 4-µm thick sections and mounted on glass slides for hematoxylin and eosin stain (H&E) and immunofluorescence (IF) analysis. A third CRAC fragment was immersed in RNAlater (Thermo Fischer Scientific, Inc.) for 48 h at 4°C. RNA was isolated using RNAeasy mini kit (Qiagen GmbH). Finally, a fourth CRAC fragment was used to isolate CSCs by culture media.

A piece of tissue (CRAC, NAT and normal colon) from each sample was washed with sterile phosphate-buffered saline (PBS; Sigma-Aldrich; Merck KGaA), immersed in 70% ethanol for 1 min (Sigma-Aldrich; Merck KGaA) and then cut into small pieces (<1 mm³) and mixed with 100 UI/ml Collagenase Type I (Invitrogen; Thermo Fisher Scientific, Inc.) diluted in 4 ml RPMI culture medium (Gibco; Thermo Fisher Scientific, Inc.), 100 µg/ml gentamicin (Gibco; Thermo Fisher Scientific, Inc.) and 2.5 µg/ml amphotericin B (Invitrogen; Thermo Fisher Scientific, Inc.). Next, the preparations were incubated for 2 h in a water bath at 37°C with a magnetic stirrer (8 stirs/min) on Cimarec hot plates (Cimarec; Thermo Fisher Scientific, Inc.). The disaggregated cells were separated from debris by sifting each of the digested preparations through a 100 µM mesh cell strainer (BD Biosciences). Then the cells were washed three times with PBS followed by resuspension in 1 ml Iscove's Modified Dulbecco's Medium (IMDM; Gibco; Thermo Fisher Scientific, Inc.) with 100 µg gentamicin/ml, 2.5 µg amphotericin B/ml (Gibco; Thermo Fisher Scientific, Inc.) and supplement 10% fetal bovine serum (FBS; Invitrogen; Thermo Fisher Scientific, Inc.). Preparations were then analyzed using the trypan blue exclusion test for cell viability. Briefly, 0.4% trypan blue solution was added to cells at room temperature. Immediately after cells without the stain (viable cells) were counted using a hemocytometer.

The evaluation of cytotoxicity in response to a mixture of first-line drugs (5FU, OXA and leucovorin) was performed with the modified ATP-CRA assay (28). Briefly, viable cells were inoculated in triplicate (20,000/50 µl) from each piece of tissue, as well as from a cell line of colorectal adenocarcinoma Colo 320DM from ATCC, in a 96-well ultralow attachment plate (Costar; Corning, Inc.) with IMDM supplemented and 10% FBS. These cells were then divided into microplates; six wells for the negative control, three wells for treated cells, and three wells contained IMDM medium as a blank, without cells or drugs. As an internal control, 105 pg ATP diluted in 100 µl IMDM medium/well was added into three wells in each microplate. Next, the mixture of the chemotherapeutic agents were used (50 µl) at a final concentration: 10 µg 5FU/ml (Teva Pharmaceutical Industries, Ltd.), 2.9 µg OXA/ml (Asofarma de México) and 0.5 µg leucovorin/ml, which represent the maximal plasma concentration of each drug. After 48 h of incubation at 37°C and 5% CO₂, the cells were lysed with 100 µl lysis buffer/well, which was then included with the CellTiter-Glo kit (Promega Corporation). The ATP content was measured in each well with the luminometer Cytation 3 (BioTek Instruments, Inc.). Luminescence was equivalent to the ATP production by cells as described elsewhere (28). The percentage of cell death (PCD) was determined as the rate of ATP luminescence reduction in the treated cultures concerning the untreated controls. Tissue with PCD ≥15.0±5.0 was classified as drug-sensitive and PCD ≤15.0 as drug-resistant. Cytotoxicity was evaluated as previously reported (28).

A total of two histological sections were obtained from each fragment. One of these sections was stained with H&E to verify the orientation of each tissue. The other section was processed for IF to visualize in situ and calculate CD44 and CD24 percentage, using monoclonal primary antibodies CD44-PE-Cy7 (cat. no. ab46793) and CD24-FITC (cat. no. ab30350; Abcam). Each antibody was diluted with PBS 1:100 v/v. Next 100 µl diluted antibody was added by slide (50 µl from each diluted antibody) and incubated at 4°C overnight. For each slide, 10 µl of mounting medium with DAPI Vectashield (Vector Laboratories, Inc.) was used to stain the nuclei of cells blue. Finally, all slides were stored at 4°C in the dark until viewed with an epifluorescence microscope (Nikon Eclipse 50i; Nikon Corporation). Colo320DM was used as the positive control and as the negative control, histological sections of the brain (negative control tissue) and colon (without primary antibody) were used.

All preparations were observed with an epifluorescence microscope at a magnification of ×400. From each of the histological sections, five microscope fields were chosen for image capture using a digital camera (Digital Sight DS-L2; Nikon Corporation). Microscope fields were chosen with the aid of the software, ENIS-Elements B12 v2.30 (Nikon Corporation). Finally, the results were reported as the mean ± standard deviation of CD44⁺ and CD24⁺ in the five microscope fields of all CRACs, NAT, normal colon and COLO 320DM. The percentage of CD44⁺, CD24⁺ and CD44⁺/CD24⁺ cells was estimated regarding the total number of cells in each histological section, by applying the following equation:

PPC (percentage of positive cells)=the number of positive cells/the total number of cells or nuclei observed by DAPI ×100.

P<0.05 by Student's t-test was considered to indicate a statistically significant difference.

Double-labeled immunohistochemistry kit confirmed the expression level of CD44 and CD24 (HRP/Green & AP/Fast Red; Abcam) according to the manufacturer's protocol, using monoclonal primary antibodies anti-CD24 (clone ALB9; 1:100) or anti-CD44 (clone F10-44-2; 1:200) and anti-CD26 (clone: polyclonal; 1:200). A total of six sensitive and six resistant tissues were evaluated to confirm a significant difference between the two groups analyzed. Slides were scanned with an Aperio AT2 Digital Pathology Scanner (AT2 Model; Leica Microsystems, Inc.), evaluating the intensity of the signal using the Sketch and Calc program (iCalc Inc., 2018; http://www.sketchandcalc.com/).

A total of 34 fresh whole samples of CRACs were submerged in 1.5 ml RNAlater for 48 h at 4°C. Afterwards, RNA was isolated using an RNAeasy mini kit and was quantified with a NanoDrop spectrophotometer (Thermo Fisher Scientific, Inc.), The RNA was kept at −80°C until ready for use. Also, RNA was quantified with RiboGreen™ (cat. no. R11490; Thermo Fisher Scientific, Inc.) using the Qubit instrument and integrity was analyzed using the Bioanalyzer RNA 6000 NanoAssay Kit, Agilent 2100 Bioanalyzer system (Agilent Technologies, Inc.). The remaining RNA was stored at −80°C until the samples were processed to construct the libraries.

Isolation and expansion of CSC were performed as described by the modified method of Cammareri et al (29). Briefly, cells from disaggregated tissue CRAC or Colo320DM were seeded in 100 µl/well (5×10⁴ cells suspended in 5 ml of medium) in 96-low adhesion plates and incubated at 37°C with 5% CO₂. Cell growth every third day was observed using an inverted microscope (VE403; Velab). During this period, 50 µl of fresh Dulbecco's Modified Eagle Medium (DMEM) F12 culture medium enriched with 6 mg glucose/ml, 1 mg NaHCO₃/ml, 5 mM 4-(2-Hydroxyethyl)-1-piperazine ethane sulfonic acid (HEPES), 2 mM glutamine, 4 µg heparin/ml, 10 ng basic fibroblast growth factor (bFGF)/ml, 20 ng epidermal growth factor (EGF)/ml, 100 µg human transferrin/ml, 25 µg insulin/ml, 9.6 µg putrescin/ml, 30 nM sodium selenite and 20 nM progesterone was added every 3–5 days to each well [all the above reagents were purchased from Sigma-Aldrich (Merck KGaA)]. Cultures were incubated at 37°C in a 5% CO₂ atmosphere for 15–20 days until typical CSC spheroids were observed.

CSC isolates (n=3 with 500,000 cell/ml) 100 µl were seeded in a slice coated w/poli-L-lysine, after fixation with Carnoy's solution for at least 48 h at 4°C. Next, quantification of CD24 and CD44 cell surface markers was performed by IF as previously mentioned. Then, five fields by color for each sample were counted by two independent observers. The graphs represented the mean obtained ± standard error. For differentiation of CSC to epithelial cells, 1×10³ CSCs were incubated in each well of the four microchambers of a Nunc Lab-Tek II Chamber Slide (Nalge Nunc International; Thermo Fisher Scientific, Inc.) at 37°C in a 5% CO₂ atmosphere for 25–30 days, with RPMI 1640 supplemented gentamicin, amphotericin B and 2% bovine serum albumin (SAB; Sigma-Aldrich; Merck KGaA), culture medium was replenished every third or fourth day. Visualization of the morphology of the cells was performed with an inverted microscope (VE403; Velab). The concentration of FBS in the culture media was increased on day 3–4 to 5% and on day 9–10 to 10% and after day 15 to 15%. The differentiated cells were fixed with methanol-acetone (1:1 v/v) and stored at 4°C until their analysis with in situ IF. Subsequently, epithelial cells were characterized using primary monoclonal antibodies 1:200 in PBS (v/v) (Abcam), anti-CK-17 (cat. no. ab51056), anti-CK-20 (cat. no. ab109111), a mixture of monoclonal antibodies against a cocktail of CK (cat. no. ab115959) signals were detected using secondary polyclonal antibodies anti-IgG [1:500 in PBS (v/v) Alexa Fluor 488; cat. no. A32731; Invitrogen; Thermo Fisher Scientific, Inc.] and anti-EGFR (cat. no. ab32562) detected secondary antibodies polyclonal anti-IgG [1:500 PBS (v/v) Alexa Fluor 660; cat. no. A-21073; Invitrogen; Thermo Fisher Scientific, Inc.]. From three randomly selected fields, images were captured at ×400 magnification. These images were merged using ImageJ version 1.52r (National Institutes of Health).

From CSCs (2,000 cells/well), resistance to 5FU/OXA was determined by ATP-CRA after 48 h of exposure to the mixture (5FU and Oxa). Then the cells (~500,000) were washed twice with PBS and incubated with fresh CSC-medium without 5FU/OXA. From these cultures (CRACs- or -Colo320), one cell fraction was selected with magnetic beads by binding anti-CD44 monoclonal antibodies. Cells were passed through an LS Column and a MidiMACS Manual Separator (Miltenyi Biotec, Inc.). Next, the eluted cell suspension was centrifuged at 500 × g for 10 min at room temperature, discarding the supernatant and resuspending the cells in 500 µl RNAlater and left at 4°C for 48 h after isolating total RNA.

RNA-seq libraries were constructed using 50–250 ng RNAs (RIN>7) with TruSeq Target RNA Expression System (Illumina, Inc.). The libraries were denatured as single-stranded DNA molecules, captured on Illumina flow cells, amplified in situ as clusters and finally sequenced for 150 cycles on an Illumina MiSeq Sequencer (Illumina, Inc.) according to the manufacturer's instructions. The cDNA library was constructed using the TruSeqRNA Access Library Prep system (Illumina, Inc.). Libraries with concentrations ≤2 nM were discarded. The size of cDNAs was determined with a 2100 Bioanalyzer system (Agilent Technologies Deutschland GmbH). The concentration of cDNAs was determined with Pico green (Thermo Fisher Scientific, Inc.), following the instructions of the manufacturer. cDNA sequencing was performed with 10 pM of the libraries with 0.1% PhiX. Single-end sequencing of multiplexed cDNA libraries was carried out on an Illumina MiSeq sequencer for 150 cycles (Reagent kit v3; Illumina, Inc.).

Data from four cDNA sequence-readings was obtained using the following equation: N=Total number of readings of each gene (TNRG)/Total number of readings per sample (TNRS) × Total number of readings of all sequenced cDNA (TNRAS). Where N denoted normalized data; TNRG, the total number of readings of each gene, TNRS, the total number of readings per sample and TNRAS, the total number of readings of all sequenced cDNA. Over 94% of reads were identified, accepting >Q30 (Q30=1 error each 1,000 bp).

To visualize the differences in gene expression associated with resistance, heatmaps were developed using GraphPad Prism 6.0 program (GraphPad Software, Inc.). The number of normalized readings for each of the samples of sensitive, resistant and control (healthy) tissues were considered and plotted. Using the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) program (https://string-db.org/), the STRING protein-protein interactions were evaluated to determine whether these genes were involved in any common metabolic pathways or functions.

Based on the results, one gene, KRT-18, with overexpression of resistance of CRAC and CSC was chosen. To confirm the results, we analyzed co-expression of KRT-18 and CD44 by IF using another 12 different paraffin-embedded CRAC only from stages III or IV (six tissues classified as drug-sensitive and drug-resistant) and six healthy tissues.

Data for chemosensitivity of primary cultures of CRAC and healthy colon samples were performed in triplicate and analyzed using Bonferroni's correction following one way ANOVA, variance analysis and Student's t-test. The percentages of the markers CD44, CD24 and CD26 were obtained in duplicate from different sections of tissues of each patient in five fields and two independent observers made analysis using Kruskal-Wallis (ANOVA-means and ranges) and Student's t-test. P<0.05 was considered to indicate a statistically significant difference.

Statistical analyses were performed using the SPSS version 20 (IBM Corp.). Results of CRAC analysis showed a PCD ≥15±5% standard deviation and a percentage ≥5 of CD44⁺ and CD24⁺ cells were considered sensitive to 5FU/OXA. From this, a heat map was constructed to show the number of normalized readings using the GraphPad Prism 8 program (GraphPad Software, Inc.). The levels of gene expression by the CSCs from sensitive and resistant CRACs were then compared with Basespace Illumina software version 4.10 (Illumina, Inc.) and Euclidean distance (absolute difference between average normal expression or sensitive CRAC and resistance CRAC). P<0.05 was considered to indicate a statistically significant difference. To identify which genes were relevant, genes found with overexpression in normal colon tissue and human mesenchymal stem cells isolated from adipose tissue (hMSC-AT) were discarded.

Finally, sets of data were analyzed by one-way ANOVA followed by the Bonferroni post hoc test. P<0.05 was considered to indicate a statistically significant difference.

To determine the threshold to define the sensitivity of the tumors, some adjacent regions from tumor and healthy tissues were evaluated with the same experimental technique. All 12/12 samples from healthy tissue (100%) had PCD ≥15.0±5.0 and were classified as drug-sensitive, Fig. 1A and B show that 22/31 (71%) CRAC-adjacent colon samples were sensitive to treatment. Then the response to 5FU/Oxa on all primary cultures (PCs) was analyzed as seen in Fig. 1C and D. This demonstrated that CRAC 22/51 (43.1%) samples were drug resistant and 29/51 (56.9%) were sensitive. Also, 14 samples (44%) from the same person (CRAC and CRAC-adjacent colon) were resistant and 17 samples (56%) were sensitive. Colo 320 was a sensitive control (92.70±00.22).

CRACs expressing CD24 and CD44 were then tested for sensitivity and resistance to 5FU and oxaliplatin. Results shown in Fig. 2A-D. Fig. 2E-G show the mean ± standard deviation of CD44⁻ and CD24-positive markers on sensitive and resistant CRACs. Briefly, the percentage of CD44⁺ cells was 4.80±2.84% and for CD24⁺ 3.85±1.48%. The mean ± standard deviation of CD44⁺/CD24⁺ cells was 2.70±2.075. By contrast, the positive percentage of CD44⁺ and CD24⁺ cells in the resistant tumor were 11.75±13.12 and 9.86±9.18%, respectively. There was a significant difference in CD44⁺ and CD24⁺ (individual markers) as sensitive and resistant tissue, as shown in Fig. 2E and F.

Total RNA (n=51) was then obtained from the CRACs cells (50-1,200 ng/µl) and the integrity analyzed. This resulted in 34 samples with a RIN >7, which were later used for the analysis of gene expression. In addition, 10 RNA samples (80–849 ng/µl) from healthy tissue had higher quality and integrity and were chosen for analysis (data not shown). Results were public and added to https://www.kaggle.com/elsangarzatrevio/quibit-and-bioanalyzer/version/1. None of the adjacent colon to CRAC was used for RNAseq.

A total of three CSC cultures from CRACs of advanced cancer tissue and from cell line Colo320 were successful and were used as a control. CSC isolates were positive to CD44 and CD24. These results demonstrated an eight times higher expression of markers after culture, as shown in Fig. 3A-F. It was also found that differentiation of CSC to epithelial cells occurs only after 15 days in medium with serum. Notably it was observed that cells acquired the ability to adhere to plastic and express EGFR (Fig. 3G), CK17 (Fig. 3H) and CK20 (Fig. 3I) markers (<80%).

CSCs were further analyzed for chemoresistance by an ATP-CRA modified assay. It was also observed that cell lines from Colo320 maintained in medium enriched CSC, permitted reduced cytotoxic activity of the drugs assayed. Results showed a decrease (Fig. 4) from 88.43±7.17 to 72.36±4.78 (P<0.05). However, CSCs isolated from tumor tissue were resistant before and after they were in the medium to enrich CSCs. After eliminating the stimulating chemotherapeutic agents, tumor cells that proliferated were observed.

A total of 66 genes were selected that were associated with chemoresistance of 5FU and oxaliplatin-based in the project (ID 40581) created with Homo sapiens (human) genome assembly GRCh37 (hg19) from Genome Reference using the Illumina platform. Next Generation Sequencing (NGS) technologies were used to assess overall gene expression profiles. Pre-validated primers for the conserved region and specific targets were selected and are shown in Table II. Gene expression levels in healthy colon samples and CSC resistance to 5FU and oxaliplatin were compared. Protein-protein interactions were also analyzed using STRING, to ascertain if the overexpression protein belonged to the same pathway or metabolic route. A heatmap of the Colo 320 line was constructed before and after treating CSC in medium (Fig. 5A); the genes that were found highly overexpressed were: Excision Repair 1, Endonuclease Non-Catalytic Subunit (ERCC1), Poly-ADP-ribosyltransferase (PARP1), TGFβ, ATP binding cassette subfamily B member 1 (ABCB1) and MYC.

Resistant and sensitive tissues had the variability of the expression genes analyzed; however, greater expression in sensitive tissues of genes involved in cell growth and differentiation was observed. By contrast, in resistant tissues, the main overexpressed genes were SOD1, MYC, PARP1, HGD, BSTG1, PKM2, GSTM1, ALDH3B and ABCB1 (Fig. 5B).

The heatmap in Fig. 6A shows the expression levels of normal stem cells human-mesenchymal stem cells-isolate from adipose tissue and CSC isolates. As observed on the expression scale, the intensity of color allowed the observation of low expression levels in blue and high expression levels in red. Results of the CSC isolates from CRACS were different in each patient; however, overexpressed genes repeated with high frequency were ribonucleoside-diphosphate reductase large subunit (RRM1), EGF, HGD and KRT-18 in colon CSC from patients with CRC (Fig. 6A). In addition, Fig. 6B represents relative expression levels of genes from CSC isolates and the tumor. It was found that RRMI, EGF, HGD and KRT-18 genes were expressed in CSC tissue.

Following the results of the present and previous studies (30,31), KRT-18 was selected as the key gene in the progression of CRAC to be validated by immunohistochemistry. Representative samples shown in Fig. 7A-F, demonstrated that co-expression of KRT-18 protein and CD44 resistance was observed by IF in the 12 tissues from the patients with CRAC. No statistical difference was found between KRT-18 and CD44 markers in sensitive and resistant tissues. Calculated mean value for KRT-18 sensitive tissue showed positive for 2.27±1.9%, while resistant showed 3.18±4.03%, meanwhile CD44 sensitive showed 2.31±2.01%, whereas resistant showed 3.32±4.02% (data not shown).